test: add benchmark regression enforcement for Phase 1

Adds AssertNoRegression method to ContentServiceBenchmarkBase:
- Loads baseline from baseline-phase0.json
- Fails test if benchmark exceeds 20% of baseline
- Logs regression check status for visibility
- Supports environment variable overrides (BENCHMARK_REGRESSION_THRESHOLD, BENCHMARK_REQUIRE_BASELINE)

Updates 10 Phase 1 CRUD benchmarks to use regression assertions:
- Save_SingleItem, Save_BatchOf100, Save_BatchOf1000
- GetById_Single, GetByIds_BatchOf100
- Delete_SingleItem, Delete_WithDescendants
- GetAncestors_DeepHierarchy, GetPagedChildren_100Items, GetPagedDescendants_DeepTree

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

tests/Umbraco.Tests.Integration/Testing/ContentServiceBenchmarkBase.cs

@@ -31,6 +31,30 @@ public abstract class ContentServiceBenchmarkBase : UmbracoIntegrationTestWithCo
 {
     private readonly List<BenchmarkResult> _results = new();
 
+    // Regression enforcement configuration
+    private const double DefaultRegressionThreshold = 20.0;
+
+    // Allow CI override via environment variable
+    private static readonly double RegressionThreshold =
+        double.TryParse(Environment.GetEnvironmentVariable("BENCHMARK_REGRESSION_THRESHOLD"), out var t)
+            ? t
+            : DefaultRegressionThreshold;
+
+    // Optional strict mode: fail if baseline is missing (useful for CI)
+    private static readonly bool RequireBaseline =
+        bool.TryParse(Environment.GetEnvironmentVariable("BENCHMARK_REQUIRE_BASELINE"), out var b) && b;
+
+    // Thread-safe lazy initialization of repository root
+    private static readonly Lazy<string> _repositoryRoot = new(FindRepositoryRoot);
+
+    // Thread-safe lazy initialization of baseline data
+    private static readonly Lazy<Dictionary<string, BenchmarkResult>> _baselineLoader =
+        new(() => LoadBaselineInternal(), LazyThreadSafetyMode.ExecutionAndPublication);
+
+    private static Dictionary<string, BenchmarkResult> Baseline => _baselineLoader.Value;
+
+    private static string BaselinePath => Path.Combine(_repositoryRoot.Value, "docs", "plans", "baseline-phase0.json");
+
     /// <summary>
     /// Records a benchmark result for later output.
     /// </summary>
@@ -118,6 +142,108 @@ public abstract class ContentServiceBenchmarkBase : UmbracoIntegrationTestWithCo
         _results.Clear();
     }
 
+    /// <summary>
+    /// Finds the repository root by searching for umbraco.sln.
+    /// </summary>
+    private static string FindRepositoryRoot()
+    {
+        var dir = new DirectoryInfo(TestContext.CurrentContext.TestDirectory);
+        while (dir != null)
+        {
+            if (File.Exists(Path.Combine(dir.FullName, "umbraco.sln")))
+            {
+                return dir.FullName;
+            }
+            dir = dir.Parent;
+        }
+        throw new InvalidOperationException(
+            $"Cannot find repository root (umbraco.sln) starting from {TestContext.CurrentContext.TestDirectory}");
+    }
+
+    /// <summary>
+    /// Records a benchmark and asserts no regression beyond the threshold.
+    /// </summary>
+    /// <param name="name">Benchmark name (must match baseline JSON key).</param>
+    /// <param name="elapsedMs">Measured elapsed time in milliseconds.</param>
+    /// <param name="itemCount">Number of items processed.</param>
+    /// <param name="thresholdPercent">Maximum allowed regression percentage (default: 20%, configurable via BENCHMARK_REGRESSION_THRESHOLD env var).</param>
+    protected void AssertNoRegression(string name, long elapsedMs, int itemCount, double thresholdPercent = -1)
+    {
+        RecordBenchmark(name, elapsedMs, itemCount);
+
+        // Use environment-configurable threshold if not explicitly specified
+        var effectiveThreshold = thresholdPercent < 0 ? RegressionThreshold : thresholdPercent;
+
+        if (Baseline.TryGetValue(name, out var baselineResult))
+        {
+            var maxAllowed = baselineResult.ElapsedMs * (1 + effectiveThreshold / 100);
+
+            if (elapsedMs > maxAllowed)
+            {
+                var regressionPct = ((double)(elapsedMs - baselineResult.ElapsedMs) / baselineResult.ElapsedMs) * 100;
+                Assert.Fail(
+                    $"Performance regression detected for '{name}': " +
+                    $"{elapsedMs}ms exceeds threshold of {maxAllowed:F0}ms " +
+                    $"(baseline: {baselineResult.ElapsedMs}ms, regression: +{regressionPct:F1}%, threshold: {effectiveThreshold}%)");
+            }
+
+            TestContext.WriteLine($"[REGRESSION_CHECK] {name}: PASS ({elapsedMs}ms <= {maxAllowed:F0}ms, baseline: {baselineResult.ElapsedMs}ms, threshold: {effectiveThreshold}%)");
+        }
+        else if (RequireBaseline)
+        {
+            Assert.Fail($"No baseline entry found for '{name}' and BENCHMARK_REQUIRE_BASELINE=true");
+        }
+        else
+        {
+            TestContext.WriteLine($"[REGRESSION_CHECK] {name}: SKIPPED (no baseline entry)");
+        }
+    }
+
+    /// <summary>
+    /// Measures, records, and asserts no regression for the given action.
+    /// </summary>
+    protected long MeasureAndAssertNoRegression(string name, int itemCount, Action action, bool skipWarmup = false, double thresholdPercent = -1)
+    {
+        // Warmup iteration (skip for destructive operations)
+        if (!skipWarmup)
+        {
+            try { action(); }
+            catch (Exception ex)
+            {
+                TestContext.WriteLine($"[WARMUP] {name} warmup failed: {ex.Message}");
+            }
+        }
+
+        var sw = Stopwatch.StartNew();
+        action();
+        sw.Stop();
+
+        AssertNoRegression(name, sw.ElapsedMilliseconds, itemCount, thresholdPercent);
+        return sw.ElapsedMilliseconds;
+    }
+
+    private static Dictionary<string, BenchmarkResult> LoadBaselineInternal()
+    {
+        if (!File.Exists(BaselinePath))
+        {
+            TestContext.WriteLine($"[BASELINE] File not found: {BaselinePath}");
+            return new Dictionary<string, BenchmarkResult>();
+        }
+
+        try
+        {
+            var json = File.ReadAllText(BaselinePath);
+            var results = JsonSerializer.Deserialize<List<BenchmarkResult>>(json) ?? new List<BenchmarkResult>();
+            TestContext.WriteLine($"[BASELINE] Loaded {results.Count} baseline entries from {BaselinePath}");
+            return results.ToDictionary(r => r.Name, r => r);
+        }
+        catch (Exception ex)
+        {
+            TestContext.WriteLine($"[BASELINE] Failed to load baseline: {ex.Message}");
+            return new Dictionary<string, BenchmarkResult>();
+        }
+    }
+
     /// <summary>
     /// Represents a single benchmark measurement.
     /// </summary>

tests/Umbraco.Tests.Integration/Umbraco.Infrastructure/Services/ContentServiceRefactoringBenchmarks.cs

@@ -65,7 +65,9 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var sw = Stopwatch.StartNew();
         ContentService.Save(content);
         sw.Stop();
-        RecordBenchmark("Save_SingleItem", sw.ElapsedMilliseconds, 1);
+
+        // Gate: Fail if >20% regression from baseline
+        AssertNoRegression("Save_SingleItem", sw.ElapsedMilliseconds, 1);
 
         Assert.That(content.Id, Is.GreaterThan(0));
     }
@@ -97,7 +99,9 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var sw = Stopwatch.StartNew();
         ContentService.Save(items);
         sw.Stop();
-        RecordBenchmark("Save_BatchOf100", sw.ElapsedMilliseconds, itemCount);
+
+        // Gate: Fail if >20% regression from baseline
+        AssertNoRegression("Save_BatchOf100", sw.ElapsedMilliseconds, itemCount);
 
         Assert.That(items.All(c => c.Id > 0), Is.True);
     }
@@ -129,7 +133,9 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var sw = Stopwatch.StartNew();
         ContentService.Save(items);
         sw.Stop();
-        RecordBenchmark("Save_BatchOf1000", sw.ElapsedMilliseconds, itemCount);
+
+        // Gate: Fail if >20% regression from baseline
+        AssertNoRegression("Save_BatchOf1000", sw.ElapsedMilliseconds, itemCount);
 
         Assert.That(items.All(c => c.Id > 0), Is.True);
     }
@@ -147,7 +153,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var id = content.Id;
 
         IContent? result = null;
-        MeasureAndRecord("GetById_Single", 1, () =>
+        MeasureAndAssertNoRegression("GetById_Single", 1, () =>
         {
             result = ContentService.GetById(id);
         });
@@ -176,7 +182,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var ids = items.Select(c => c.Id).ToList();
 
         IEnumerable<IContent>? results = null;
-        MeasureAndRecord("GetByIds_BatchOf100", itemCount, () =>
+        MeasureAndAssertNoRegression("GetByIds_BatchOf100", itemCount, () =>
         {
             results = ContentService.GetByIds(ids);
         });
@@ -194,7 +200,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var content = ContentBuilder.CreateSimpleContent(ContentType, "DeleteTest", -1);
         ContentService.Save(content);
 
-        MeasureAndRecord("Delete_SingleItem", 1, () =>
+        MeasureAndAssertNoRegression("Delete_SingleItem", 1, () =>
         {
             ContentService.Delete(content);
         }, skipWarmup: true); // Destructive operation - cannot repeat
@@ -220,7 +226,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
             ContentService.Save(child);
         }
 
-        MeasureAndRecord("Delete_WithDescendants", childCount + 1, () =>
+        MeasureAndAssertNoRegression("Delete_WithDescendants", childCount + 1, () =>
         {
             ContentService.Delete(parent);
         }, skipWarmup: true); // Destructive operation - cannot repeat
@@ -253,7 +259,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         IEnumerable<IContent>? results = null;
         long totalRecords = 0;
 
-        MeasureAndRecord("GetPagedChildren_100Items", childCount, () =>
+        MeasureAndAssertNoRegression("GetPagedChildren_100Items", childCount, () =>
         {
             results = ContentService.GetPagedChildren(parent.Id, 0, 100, out totalRecords);
         });
@@ -293,7 +299,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         IEnumerable<IContent>? results = null;
         long totalRecords = 0;
 
-        MeasureAndRecord("GetPagedDescendants_DeepTree", 300, () =>
+        MeasureAndAssertNoRegression("GetPagedDescendants_DeepTree", 300, () =>
         {
             results = ContentService.GetPagedDescendants(root.Id, 0, 1000, out totalRecords);
         });
@@ -324,7 +330,7 @@ internal sealed class ContentServiceRefactoringBenchmarks : ContentServiceBenchm
         var deepestId = current.Id;
 
         IEnumerable<IContent>? results = null;
-        MeasureAndRecord("GetAncestors_DeepHierarchy", depth, () =>
+        MeasureAndAssertNoRegression("GetAncestors_DeepHierarchy", depth, () =>
         {
             results = ContentService.GetAncestors(deepestId);
         });