fix: shorten interrupted task redelivery

[suguanYang]

Summary

• reduce the Redis broker visibility timeout from 12 hours to 75 minutes
• keep the task hard limit, visibility timeout, and stale-job expiry as one validated ordering
• reject environment overrides that would expire a processing job before Redis can redeliver it
• add worker contract coverage for the valid runtime values and invalid override

Why

A hard-stopped late-acknowledged task currently remains invisible for 12 hours, while the stale-job sweeper marks processing jobs failed after 4 hours. The terminal job gate then prevents the eventual redelivery from restarting the job. The new ordering is:

3600s task hard limit < 4500s visibility timeout < 14400s processing expiry

This keeps the visibility timeout above the maximum task runtime while ensuring hard-stop redelivery occurs before stale-job expiry. Redelivery restarts processing from the beginning; it is not checkpoint resume.

Verification

uv run --directory /home/suguan/github.com/ontosAI/knowhere pytest -q apps/worker/tests/contract/test_worker_shutdown_contract.py apps/worker/tests/contract/test_worker_bootstrap_contract.py
5 passed

uv run --directory /home/suguan/github.com/ontosAI/knowhere ruff check packages/shared-python/shared/core/config/app.py packages/shared-python/shared/core/config/celery.py packages/shared-python/shared/core/celery_app.py apps/worker/tests/contract/test_worker_shutdown_contract.py
All checks passed!

Related to Ontos-AI/knowhere-api-infra#22.