feat(goose): complete v1.49 phase3 closeout gates and context fusion plan

Expand Goose v1.49 smoke coverage (memory chat, portal resume, page e2e,
multiturn provider), add canary memory policy lock, refresh baselines, and
ignore one-off evidence artifacts. Document headroom-based context runtime
fusion plan; include auth, scheduled-task, and wechat intent fixes on branch.

Co-authored-by: Cursor <cursoragent@cursor.com>

memory-v2-runtime.mjs

@@ -1,4 +1,5 @@
 import path from 'node:path';
+import { applyGooseCanaryMemoryPolicy } from './goose-canary-memory-policy.mjs';
 import { pathToFileURL } from 'node:url';
 import { createLangGraphHttpClient, createLangGraphMemoryBackend } from './memory-v2-langgraph.mjs';
 import { createLegacyMemoryBackend, createMemoryV2, resolveMemoryV2Policy } from './memory-v2.mjs';
@@ -677,6 +678,12 @@ export async function createManagedMemoryV2Runtime({
 
   async function ensureRuntime() {
     const state = await loadRuntimeState();
+    state.effectiveEnv = applyGooseCanaryMemoryPolicy(state.effectiveEnv, env);
+    const canaryMode = String(env.TKMIND_GOOSE_CANARY ?? 'off').trim().toLowerCase();
+    if (canaryMode === 'all' || canaryMode === 'users') {
+      state.source = `goose-canary:${state.source}`;
+    }
+    state.fingerprint = `${state.fingerprint}:goose-canary=${canaryMode}`;
     if (activeRuntime && activeFingerprint === state.fingerprint) {
       activeMeta = state;
       return activeRuntime;