Improve MLflow artifact upload diagnostics and error handling

[Copilot]

Addresses diagnostic and error handling gaps in the MLflow artifact upload implementation based on review feedback.

Changes

• Consistent failure tracking: Reset consecutive_failures counter on successful trace file uploads (matching log file behavior)
• Better diagnostics in upload_mlflow_artifacts: Separate MLflow-not-requested from wrong-rank conditions to provide clearer error messages when artifact upload fails silently
• Additional warning: Detect when MLflow is requested but writer isn't initialized to avoid silent failures

Example

# Before: Silent failure if MLflow writer is None for any reason
if mlflow_writer is None:
    return None

# After: Distinguish between expected and unexpected None cases
if mlflow_run_name is None:
    return None  # MLflow not requested - expected
    
if args.rank != expected_rank:
    debug_rank_all("WARNING: called from wrong rank...")  # Wrong rank - expected
    return None
    
if mlflow_writer is None:
    debug_rank_0("WARNING: writer not initialized...")  # Unexpected - helps debugging

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[Copilot]

Copilot wasn't able to review any files in this pull request.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.