Keep the restart_workers monitor alive on transient errors (and drop debug prints)

[lollinng]

Problem

The restart_workers watchdog in _start_worker_monitoring.monitor() wraps its entire while not self._shutdown_event.is_set() loop in a single try/except:

def monitor():
    try:
        while not self._shutdown_event.is_set():
            ...           # map dead workers, relaunch them
    except Exception as e:
        print(e)

So if any single iteration raises — a transient worker→API mapping error, a failed launch_single_inference_worker, etc. — the exception breaks out of the while loop, the monitor thread exits, and restart_workers self-healing is permanently disabled for the rest of the server's lifetime. The loop also shipped three leftover debug prints (misspelled [monoriting]).

Fix

Move the try/except inside the loop so a single bad iteration is logged (logger.exception) and the watchdog keeps monitoring. Replace the [monoriting] debug prints with logger calls. The intentional return on restart_workers=False (graceful shutdown) is preserved.

Verification

Modeled both control-flow shapes with a body that raises once then succeeds:

OLD: ran 0 iterations then the monitor thread DIED (self-healing disabled)
NEW: ran 3 iterations — survived the transient error and kept monitoring