- Add visibility_timeout (12h) to prevent tasks from being redelivered
  while still running
- Enable task_acks_late and task_reject_on_worker_lost to ensure tasks
  are requeued if a worker crashes
- Set worker_prefetch_multiplier=1 to prevent task hoarding
- Add broker_heartbeat for faster connection problem detection
- Configure socket timeouts and retry_on_timeout for Redis connections
- Add exponential backoff retry (3 retries) for Redis client
- Add health_check_interval for periodic connection validation
- Add configurable task time limits (24h hard, 23h soft)
- Fix task_track_started from tuple (True,) to boolean True

New environment variables for configuration:
- CELERY_VISIBILITY_TIMEOUT, CELERY_BROKER_HEARTBEAT
- CELERY_TASK_TIME_LIMIT, CELERY_TASK_SOFT_TIME_LIMIT
- CELERY_BROKER_POOL_LIMIT, CELERY_REDIS_MAX_CONNECTIONS
- REDIS_SOCKET_TIMEOUT, REDIS_SOCKET_CONNECT_TIMEOUT
- REDIS_HEALTH_CHECK_INTERVAL

AI-assisted: Claude Code

Signed-off-by: Christian Berendt <berendt@osism.tech>