[Feature] Add modular tracking interface with MLflow backend

[mouad-hpc]

Summary

• Adds TrackingBackend ABC and TrackingManager for pluggable logging backends
• Adds MLflow as a new tracking backend alongside wandb and tensorboard
• Refactors logging_utils.py to delegate to the manager instead of hardcoded conditionals
• New CLI flags: --use-mlflow, --mlflow-tracking-uri, --mlflow-experiment-name, --mlflow-run-name
• MLflow is an optional dependency (pip install slime[mlflow])

Files changed

• slime/utils/tracking.py (new) - shared interface + adapters + registry
• slime/utils/mlflow_utils.py (new) - MLflow backend implementation
• slime/utils/logging_utils.py - uses TrackingManager internally, same public API
• slime/utils/arguments.py - MLflow CLI flags
• setup.py - mlflow added as optional extra
• requirements.txt - mlflow removed from core deps

Test plan

• Verified e2e functionality on 8xH100 GPU cluster
• Full training run with --use-mlflow logging metrics end-to-end
• Verify secondary rank attachment via mlflow_run_id
• Verify no regression when mlflow is not installed

[mouad-hpc]

@zhuzilin For review