feat(train): add validation metrics and distributed support

[ha405]

This PR extends the validation metrics functionality (precision, recall, F1-score) to the train.py script.

Changes:

• The validate function within train.py now supports the --metrics-avg flag.
• Implemented torch.distributed.all_gather to correctly collect predictions and targets from all GPUs before calculating metrics on the primary process.
• The feature remains a soft dependency on scikit-learn and is disabled by default.

This ensures that users can get these more detailed metrics during training, even in a multi-GPU environment.