Fix reminder delivery retry monitoring

[benbenlijie]

Summary

Implements a production-oriented retry and monitoring layer for reminder delivery jobs.

This PR adds:

• persisted reminder job execution metadata: retry count, next retry time, last error, status, timestamps
• exponential backoff retry scheduling with failed/dead-letter handling
• a reminder job service for due job processing, manual retry, and monitoring stats
• monitoring endpoint for reminder jobs
• manual retry endpoint for a specific reminder
• OpenAPI/schema updates for the new retry/monitoring behavior
• backend tests for retry success, retry backoff, max retry failure, monitoring, and manual retry flows

Fixes #130

Verification

Run from packages/backend:

. .venv/bin/activate
python -m pytest -q
python -m flake8 app tests

Results:

28 passed in 5.18s
flake8 passed

Notes

To keep the test suite self-contained, tests use a small FakeRedis fixture so reminder job behavior can be verified without requiring a live Redis service.