[TO_REVIEW] Enhancements to DomainAwareNet and Scorers to handle allow_source arg

[YanisLalou]

The most important changes include adding support for source domains in various methods and updating the scoring mechanism to accommodate these changes.

Enhancements to DomainAwareNet:

• Added allow_source parameter to predict, predict_proba, and score methods to support source domains. (skada/deep/base.py) [1] [2] [3]
• Modified check_X_domain usage in predict_proba and score methods to handle allow_source. (skada/deep/base.py) [1] [2]

Updates to testing and metrics:

• Updated tests to include allow_source parameter in predictions. (skada/deep/tests/test_deep_scorer.py)
• Removed commented-out scorer in tests. (skada/deep/tests/test_deep_scorer.py)
• Updated _score method in metrics.py to handle allow_source parameter and removed outdated code. (skada/metrics.py) [1] [2] [3]
• Added allow_source parameter to various scoring functions and ensured compatibility with deep estimators. (skada/metrics.py) [1] [2] [3] [4]

[antoinecollas]

You should add a test to check an error is raised when allow_source = False and positive sample_domain is provided to a deep model.