fix: repair full-suite failures after SQLite-to-PostgreSQL cutover (#2086)

## Summary

Fixes all deterministic full-suite (non-blocking) CI failures on `main`
caused by the SQLite-to-PostgreSQL cutover (VAL-REMOVAL-005).

## Changes

### i18n Key Parity (5 locale files)
- Added missing `taskPopupsBoardListOnly` +
`taskPopupsBoardListOnlyHelp` keys (empty strings per convention) to
zh-CN, zh-TW, fr, es, ko `app.json`

### Dashboard Curated-Gate Guard (`scripts/lib/test-quarantine.json`)
- Repaired "mirror drift": 16 dashboard test files were quarantined in
`vitest.config.ts` but never added to the quarantine ledger. Added all
16 with failing run URLs and `quarantinedAt` dates.

### Line-Count Audit CI Cache (`.github/workflows/full-suite.yml`)
- Removed `skip-install: "true"` from `line-count-audit` job —
`setup-node@v5` with `cache: pnpm` failed post-step because no
`node_modules` existed to cache.

### Engine Slow Tier — Full PG Migration
- **CI**: Added PostgreSQL service container to `test-slow` job (same
config as `test-shards`)
- **`_helpers.ts`**: Migrated `makeReliabilityFixture()` from removed
SQLite `Database.init()` to PG-backed `TaskStore`:
  - Added `probeTcpReachable()` (TCP probe, copied from shared harness)
  - Added `hasPg` export (uses TCP probe, not env-var guess)
- Added `adminExecAsync()` (`Promise.withResolvers`, psql via
`PG_TEST_URL_BASE`)
- Added `createPgLayer()` (fresh PG database + schema baseline +
`AsyncDataLayer`)
  - Updated cleanup: `await store.close()`, close layer, drop database
- **Slow test**: Migrated 24 sync SQLite API calls to async PG APIs:
- `store.getRunAuditEvents()` → `await auditEvents(store, ...)` via
exported `queryRunAuditEvents`
- `store.getDatabase().prepare(...)` → Drizzle queries via
`store.getAsyncLayer()!.db`
- **Core exports**: Added `queryRunAuditEvents` from `async-audit.ts`
and `eq as drizzleEq` from `drizzle-orm`
- **22 reliability test files**: Added `hasPg` guards so tests skip
locally when PG is unavailable

### Shard 3 — PG Test Auth Bug (18 postgres test files)
- Replaced `psql -U ${process.env.USER ?? "postgres"}` with `psql
"${PG_TEST_URL_BASE}/postgres"` connection string. On GitHub Actions,
`process.env.USER` is `'runner'`, not `'postgres'`, causing auth
failure.

### Shard 3 — Removed Function Tests (`mesh-task-replication.test.ts`)
- Deleted 3 tests for functions intentionally removed in PostgresCutover
(`buildMeshReplicatedTaskCreatePayload`, `toReplicatedCreateInput`,
`taskMatchesReplicatedCreate`). Kept `buildBootstrapPrompt` test.

### Shard 3 — Store Thinking Levels (`store-thinking-levels.test.ts`)
- Migrated from removed SQLite path to PG-backed
`createTaskStoreForTest` + `pgDescribe`.

## Verification

| Check | Result |
|---|---|
| Merge gate (`pnpm test:gate`) | ✅ 294 + 99 + 63 = 456 passed |
| Engine slow tier (22 tests) | ✅ 22/22 passed |
| i18n parity tests | ✅ 7 passed |
| mesh-task-replication | ✅ 1 passed |
| PG data-layer | ✅ 14 passed |
| PG taskstore-lifecycle | ✅ 16 passed |
| store-thinking-levels | ✅ 1 passed |
| Dashboard curated-gate | ✅ passes |
| Typecheck (engine + core) | ✅ clean |
| Lint | ✅ exit 0 |

## Parked (not in scope)

- **Shards 1/2 timeout**: Engine test suite exceeds CI time budget.
Pre-existing, unrelated to these fixes.
- **2 latent PG files** (`chat-store-content-search-edit`,
`satellite-db-injected-stores`): Surface a separate pre-existing schema
baseline gap. Out of scope.

.github/workflows/full-suite.yml

@@ -197,8 +197,6 @@ jobs:
 
       - name: Setup Node.js and pnpm
         uses: ./.github/actions/setup-node-pnpm
-        with:
-          skip-install: "true"
 
       # FNXC:TestInfrastructure 2026-06-21-10:26:
       # Keep line-count drift visible in automated post-merge signal without restoring it to the blocking PR gate.
@@ -214,6 +212,27 @@ jobs:
   test-slow:
     name: Engine slow tier
     runs-on: ubuntu-latest
+    # FNXC:FixPgTestsAndCi 2026-07-14-00:00:
+    # Provision a PostgreSQL service container so reliability-interaction test
+    # helpers (which now require a PG-backed TaskStore after SQLite removal
+    # VAL-REMOVAL-005) can run in the engine-slow tier too.
+    services:
+      postgres:
+        image: postgres:15
+        env:
+          POSTGRES_USER: postgres
+          POSTGRES_PASSWORD: postgres
+          POSTGRES_DB: postgres
+        ports:
+          - 5432:5432
+        options: >-
+          --health-cmd "pg_isready -h localhost -p 5432 -U postgres"
+          --health-interval 5s
+          --health-timeout 5s
+          --health-retries 10
+    env:
+      FUSION_PG_TEST_URL_BASE: "postgresql://postgres:postgres@localhost:5432"
+      PGPASSWORD: "postgres"
     # Real-git slow suites; same backstop rationale as test-shards. Sits above
     # the L2 per-invocation ceiling so the watchdog fires first on a single hang.
     timeout-minutes: 60