Collect performance statistics and progress when training models

[ashwin153]

• Might be nice to display in the UI (we tried n features, it took y seconds, etc.)
• Would give us some quantitative ways to measure performance improvements

[ashwin153]

with streamlit.status(...):
    handler_id = loguru.add(...)
    try:
        train_model()
    finally:
        loguru.remove(handler_id)

• Register a custom logging.Handler that emits logs using streamlit.write