fix(notification_router): handle partial failures in batch_load_and_notify

[bodymindarts]

Summary

• When repo.find_all() fails for an entire batch, fall back to individual find_by_id() calls so a failure for one job does not block notifications for the rest
• Each individual failure is logged with a warning but processing continues for the remaining jobs
• Successful jobs still get their waiters notified immediately

Addresses Bugbot observation 2 from #76.

Test plan

• Existing integration tests pass (happy path unchanged — find_all still used as fast path)
• If find_all encounters a transient DB error, individual loads proceed and unrelated waiters are notified

🤖 Generated with Claude Code

---

Note

Medium Risk
Changes notification delivery behavior in the terminal-state path; a find_all failure now triggers N individual DB reads, which could affect load/latency and retry semantics under DB errors.

Overview
Ensures terminal job notifications aren’t blocked when the batched repo.find_all() call fails in batch_load_and_notify.

On batch failure, it now logs the fallback and attempts load_and_notify per job ID, allowing unaffected jobs’ waiters to be notified immediately instead of waiting for the next sweep.

^{Written by Cursor Bugbot for commit a7f9288. This will update automatically on new commits. Configure here.}

[bodymindarts]

Closing per review — the find_all fallback adds unnecessary complexity for a case that shouldn't happen in practice.