Telemetry thread exceptions are silently swallowed

[joelsmithTT]

Issue

When telemetry extraction fails (e.g. during chip reset), the telemetry update thread dies silently without showing any error message or traceback. This makes debugging very difficult.

Current behavior: Heartbeat stops, telemetry freezes, no error shown
Expected behavior: Exception and traceback displayed when telemetry fails

Reproduction

1. Start tt-smi and navigate to the Telemetry tab
2. In another terminal, reset a chip: tt-smi -r 0
3. Observe: heartbeat stops updating, but no error message appears

Root Cause

The telemetry thread uses threading.Thread(daemon=True) which silently terminates on exceptions.

Current code (tt_smi/tt_smi.py, line ~574):

thread = threading.Thread(target=update_telem, name="telem_thread")
thread.setDaemon(True)  # Daemon threads die silently on exception
thread.start()

Suggested Fix

Use Textual's Worker API, which properly displays exceptions:

# Simple fix: add thread=True parameter
worker = self.run_worker(
    update_telem, 
    thread=True,
    exit_on_error=True,  # Show traceback and exit (helpful for debugging)
    name="telem_worker"
)
TELEM_THREADS.append(worker)

For production use, set exit_on_error=False to continue on transient errors (though you'll want error handling inside update_telem() in that case).

Reference

From Textual docs:

A Worker that encounters an exception will, by default, display a traceback and exit the app. This behavior is designed to help you quickly identify issues during development.

[joelsmithTT]

A bit more context: I was trying to understand why the telemetry heartbeat freezes when looking into tt-burnin vs tt-smi as part of tenstorrent/luwen#117 testing. The Suggested Fix above was AI-generated and worked enough for me to validate my hypothesis (chip reset breaks an active tt-smi's telemetry), although it does not appear to be a complete solution (I get AttributeError: 'Worker' object has no attribute 'join' when quitting a non-interrupted-by-reset tt-smi).