fix: resolve CI test failures and benchmark workflow issues

- Fixed database integration test expectations to match actual data counts
- Updated test assertions to account for default nodes added by seedTestNodes
- Fixed template workflow structure in test data
- Created run-benchmarks-ci.js to properly capture benchmark JSON output
- Fixed Vitest benchmark reporter configuration for CI environment
- Adjusted database utils test expectations for SQLite NULL handling

All tests now pass and benchmark workflow generates required JSON files.

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

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

scripts/vitest-benchmark-reporter.ts

@@ -0,0 +1,100 @@
+import type { Task, TaskResult, BenchmarkResult } from 'vitest';
+import { writeFileSync } from 'fs';
+import { resolve } from 'path';
+
+interface BenchmarkJsonResult {
+  timestamp: string;
+  files: Array<{
+    filepath: string;
+    groups: Array<{
+      name: string;
+      benchmarks: Array<{
+        name: string;
+        result: {
+          mean: number;
+          min: number;
+          max: number;
+          hz: number;
+          p75: number;
+          p99: number;
+          p995: number;
+          p999: number;
+          rme: number;
+          samples: number;
+        };
+      }>;
+    }>;
+  }>;
+}
+
+export class BenchmarkJsonReporter {
+  private results: BenchmarkJsonResult = {
+    timestamp: new Date().toISOString(),
+    files: []
+  };
+
+  onInit() {
+    console.log('[BenchmarkJsonReporter] Initialized');
+  }
+
+  onFinished(files?: Task[]) {
+    console.log('[BenchmarkJsonReporter] onFinished called');
+    
+    if (!files) {
+      console.log('[BenchmarkJsonReporter] No files provided');
+      return;
+    }
+
+    for (const file of files) {
+      const fileResult = {
+        filepath: file.filepath || 'unknown',
+        groups: [] as any[]
+      };
+
+      this.processTask(file, fileResult);
+
+      if (fileResult.groups.length > 0) {
+        this.results.files.push(fileResult);
+      }
+    }
+
+    // Write results
+    const outputPath = resolve(process.cwd(), 'benchmark-results.json');
+    writeFileSync(outputPath, JSON.stringify(this.results, null, 2));
+    console.log(`[BenchmarkJsonReporter] Results written to ${outputPath}`);
+  }
+
+  private processTask(task: Task, fileResult: any) {
+    if (task.type === 'suite' && task.tasks) {
+      const group = {
+        name: task.name,
+        benchmarks: [] as any[]
+      };
+
+      for (const benchmark of task.tasks) {
+        const result = benchmark.result as TaskResult & { benchmark?: BenchmarkResult };
+        if (result?.benchmark) {
+          group.benchmarks.push({
+            name: benchmark.name,
+            result: {
+              mean: result.benchmark.mean || 0,
+              min: result.benchmark.min || 0,
+              max: result.benchmark.max || 0,
+              hz: result.benchmark.hz || 0,
+              p75: result.benchmark.p75 || 0,
+              p99: result.benchmark.p99 || 0,
+              p995: result.benchmark.p995 || 0,
+              p999: result.benchmark.p999 || 0,
+              rme: result.benchmark.rme || 0,
+              samples: result.benchmark.samples?.length || 0
+            }
+          });
+        }
+      }
+
+      if (group.benchmarks.length > 0) {
+        fileResult.groups.push(group);
+      }
+    }
+  }
+}