What does this PR do?

Fixes the TODO in loss_for_object_detection.py:196 - mask loss now properly ignores padding areas.

When batching images of different sizes, nested_tensor_from_tensor_list pads smaller images. Previously, padding areas were included in dice_loss and sigmoid_focal_loss computation. Now we use the valid mask to exclude them.

Changes:

• Add optional valid_mask parameter to dice_loss and sigmoid_focal_loss
• Use valid mask in loss_masks method (both loss_for_object_detection.py and loss_rt_detr.py)
• AMP compatible (preserves input dtype)
• Backward compatible (valid_mask=None by default)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

Who can review?

@molbap @ArthurZucker