fix(import): make performance baselines trustworthy

Run benchmarks in isolated GC-enabled workers, report sparse RSS samples accurately, and hash canonical imported output. Require the Rust Native WeFlow kernel and reject any measured import that falls back to TypeScript so timing and memory baselines remain comparable.

packages/node-runtime/src/import/auto-importer.test.ts

@@ -78,8 +78,8 @@ function createDeps(options?: {
             messageBatchCount: 0,
             messageTransactionCount: 0,
             rssStartMb: 0,
-            rssPeakMb: 0,
-            rssDeltaMb: 0,
+            rssSampledPeakMb: 0,
+            rssSampledDeltaMb: 0,
           },
         },
       }

packages/node-runtime/src/import/streaming-importer.test.ts

@@ -89,8 +89,8 @@ function assertPerformanceDiagnostics(result: StreamImportResult): void {
   assert.ok(diagnostics.timings.totalMs >= diagnostics.timings.messageWriteMs)
   assert.ok(diagnostics.messageBatchCount > 0)
   assert.ok(diagnostics.messageTransactionCount > 0)
-  assert.ok(diagnostics.rssPeakMb >= diagnostics.rssStartMb)
-  assert.ok(diagnostics.rssDeltaMb >= 0)
+  assert.ok(diagnostics.rssSampledPeakMb >= diagnostics.rssStartMb)
+  assert.ok(diagnostics.rssSampledDeltaMb >= 0)
 }
 
 function writeChunkedShuakamiQqExport(root: string): string {

packages/node-runtime/src/import/streaming-importer.ts

@@ -71,8 +71,9 @@ export interface ImportPerformanceDiagnostics {
   messageBatchCount: number
   messageTransactionCount: number
   rssStartMb: number
-  rssPeakMb: number
-  rssDeltaMb: number
+  /** Highest RSS observed at import stage and parser batch boundaries; not a continuous process peak. */
+  rssSampledPeakMb: number
+  rssSampledDeltaMb: number
 }
 
 export interface StreamImportResult {
@@ -843,8 +844,8 @@ async function streamImportSingle(
       messageBatchCount: callbackStats.onMessageBatchCalls,
       messageTransactionCount,
       rssStartMb: rssStartBytes / bytesPerMegabyte,
-      rssPeakMb: rssPeakBytes / bytesPerMegabyte,
-      rssDeltaMb: (rssPeakBytes - rssStartBytes) / bytesPerMegabyte,
+      rssSampledPeakMb: rssPeakBytes / bytesPerMegabyte,
+      rssSampledDeltaMb: (rssPeakBytes - rssStartBytes) / bytesPerMegabyte,
     },
   }