All four tasks done. Here's a summary of every change made:

---
S1 — Persist scraped snapshot

fk_scraper.py:
- Added persist_callback: Callable[[ParentalLeaveFetch], Awaitable[None]] | None field to _SessionState
- Added persist_callback parameter to start_session()
- Updated _do_extract(): after successful extraction, awaits the callback; on callback failure sets status=error instead of done

routers/fetch_data.py:
- Added parental_leave_repo: ParentalLeaveRepositoryDep to start_fetch
- Created a _persist closure capturing the repo and passed it to start_session() — the shared app.state.db_connection is safe to use from the background coroutine (aiosqlite serialises all operations through its
internal queue)

S2 — Read contract (Option A)

No backend changes needed. The existing GET /api/parental-leave/{personal_data_uuid}/{child_id} route (fetch_data.py:81) already calls get_latest_by_parent which orders by fetched_at DESC with no source filter —
it returns the latest snapshot regardless of whether it's scraped or manual. The junior should point api.ts to exactly that path; both IDs come from GET /api/auth/context.

S3 — Scraper enabled

- backend/.env created with APP_SCRAPER_ENABLED=true (gitignored — enables the scraper for the running dev app)
- backend/.env.example created as documentation
- Stale "DO NOT SHIP / pending legal review" docstring in fk_scraper.py replaced with "legal cleared 2026-06" + guardrail summary
- _SCRAPER_DISABLED 501 message updated (no longer says "pending legal review")
- config.py comment updated to reference APP_SCRAPER_ENABLED=true and note playwright install chromium requirement

Tests (S1 acceptance)

3 new tests in test_fk_scraper.py:
- test_do_extract_calls_persist_callback_on_success — unit: callback is invoked, status is done, result is set
- test_do_extract_sets_error_when_persist_callback_fails — unit: callback failure → status=error, not done
- TestPersistenceIntegration::test_scraped_row_is_readable_after_session_done — integration: real SQLite repo, row persisted, get_latest_by_parent returns the scraped row

Full suite: 178 passed (pre-existing 1 failure in test_tips_service.py unchanged).