test: add frontend unit suite + fix mixed local/UTC timestamps

Two CODE_AUDIT items in one session.

§3.2 — Frontend regression net (js/tests/, 35 tests, node:test, zero deps):
- harness.js loads app.js (monofile, no exports) into a node:vm with browser
  globals stubbed, surfacing internals via an export epilogue.
- crypto: deriveKeyAndVerifier (AES key == raw PBKDF2, cross-checked vs Node
  pbkdf2Sync), legacy-vs-v2 verifier decoupling, encrypt/decrypt round-trip,
  IV uniqueness, AEAD tamper/wrong-key.
- csv: parseCSV tokenizer, findColumn heuristics, Bitwarden/KeePass mapping.
- merge: applyRemoteSnapshot add/update/skip (LWW), tombstone delete,
  resurrection arbitration (both NaN branches), local-tombstone veto,
  additive folder merge. Only api() is stubbed; loadEntries/encryptImportEntry
  run for real.
- Wired as a build gate in BuildAssets.ps1 (after node --check, bypass
  PM_SKIP_TESTS=1).

§2.2 — Unify timestamps on UTC:
- Entry created_at/updated_at were written via Delphi FormatDateTime(Now)
  = LOCAL, while deleted_at/tombstones use SQLite CURRENT_TIMESTAMP = UTC.
  The tombstone-resurrection arbitration compared the two zones, skewing by
  the machine's UTC offset even single-device.
- Add NowUTC/NowUTCStr to PM.Database, swap in at every entry/attachment
  write site (Entries create/update/bulk, Attachments POST echo).
- No JS change needed: arbitration now compares same-zone values.
- Existing rows self-heal on next edit (no destructive migration).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

js/tests/README.md

@@ -0,0 +1,58 @@
+# Frontend unit tests
+
+Regression net for the highest-risk pure/near-pure logic in `js/app.js`:
+crypto round-trip, KDF/verifier derivation, CSV import parsing, and the
+sync-merge / tombstone-resurrection arbitration. Addresses `CODE_AUDIT.md`
+§3.2 (aucun test automatisé).
+
+## Running
+
+```
+npm test
+# or directly:
+node --test "js/tests/**/*.test.js"
+```
+
+Zero dependencies — uses the Node built-in test runner (`node:test`) and
+`webcrypto`. Requires Node ≥ 18 (developed on v24). Runs in ~1.7 s.
+
+## How it works — `harness.js`
+
+`app.js` is a ~12k-line browser monofile with **no module exports** and one
+top-level side effect (a `DOMContentLoaded` listener). The harness loads the
+file's source into a `node:vm` context with browser globals stubbed
+(`crypto`, `localStorage`, `document`, `location`, …) so `init()` never
+fires, then appends an export epilogue that surfaces the internals on
+`globalThis.__test`.
+
+Two gotchas the harness works around, both documented inline:
+
+- **`const`/`let` don't attach to the vm global.** Top-level `function`/`var`
+  declarations become properties of the context global, but `const state`,
+  `const HASH_ALGO_V2`, etc. do not — hence the explicit export epilogue.
+- **Cross-realm prototypes.** Values returned from the sandbox carry the
+  sandbox realm's prototypes, so `assert.deepStrictEqual` trips on the
+  prototype check. Structural comparisons normalize through JSON first
+  (see `eqDeep` in `csv.test.js`).
+
+The merge tests stub only the `api()` seam (a `function` declaration →
+overridable global property) with an in-memory fake server; `loadEntries`,
+`loadFolders`, and `encryptImportEntry` all run for real against it — so the
+tests exercise the actual pull→merge path, not a re-implementation.
+
+## Suites
+
+| File | Covers |
+|---|---|
+| `crypto.test.js` | `deriveKeyAndVerifier` (AES key == raw PBKDF2, cross-checked vs Node's `pbkdf2Sync`), legacy vs `-v2` verifier decoupling, `encryptPwd`/`decryptPwd` round-trip, IV uniqueness, AEAD tamper/wrong-key → `[ERROR]` |
+| `csv.test.js` | `parseCSV` tokenizer (quotes, escaped `""`, CRLF, trailing field), `findColumn` header heuristics, `parseEntriesFromCSV` for Bitwarden/KeePass shapes, note-vs-login classification |
+| `merge.test.js` | `applyRemoteSnapshot`: add/update/skip (last-write-wins), tombstone delete, resurrection arbitration (both `NaN` branches), local-tombstone veto, additive folder merge |
+
+## Notes on encoded behavior
+
+`merge.test.js` locks in one deliberately-asymmetric behavior: an unparseable
+**local** `updated_at` favours KEEP (resurrection wins), but an unparseable
+**remote** `deleted_at` still applies the delete. In production `deleted_at`
+is always server-formatted (parseable), so this only matters for a corrupted
+remote snapshot. If that arbitration is ever changed, the two
+`unparseable …` tests are where to update the expectation.