fix: repair sprint completion and add QA pass

[numnx]

What changed

• Replaced the generic sprint status patch behind Mark Completed with a runtime-aware manual completion endpoint.
• Added Mark QA Pass to sprint card and ledger action menus, including pending, active-review, and already-passed states.
• Persisted manual sprint QA passes as durable QA review runs and resolved matching sprint-scoped QA handoffs.
• Added backend service, repository, route, dashboard, and failure-path coverage.
• Updated canonical and published dashboard, QA, execution-control, and HTTP API documentation.

Why

The previous completion button changed only the sprint record. When a sprint run was still active, its runtime projection could immediately overwrite the manual completed status, making the button appear broken. Manual QA acceptance also had no sprint-page control or durable QA verdict.

User impact

• Mark Completed now safely stops active sprint runtime work before persisting completion, so the state remains stable.
• Mark QA Pass records an explicit manual pass after human review without dismissing task QA or unrelated attention.
• QA pass is unavailable during an active sprint QA review and disappears after a passing verdict.

Validation

• pnpm run lint
• pnpm run test:backend — 437 files passed, 5,490 tests passed, 7 skipped
• pnpm run test:dashboard — 241 files passed, 2,152 tests passed, 2 skipped
• pnpm run build
• Dashboard /sprints returned HTTP 200 on port 4444

Risk and rollback

Risk is limited to explicit manual sprint actions. Runtime cleanup is fail-closed: if active-run cancellation fails, completion is not persisted. Roll back by reverting this commit, which restores the previous generic status patch and removes the manual QA endpoint/action.