[MRG] Handle edge case Mixvalscorer

[YanisLalou]

This pull request includes changes to handle cases where there are no indices for intra-cluster or inter-cluster evaluations in the MixValScorer, and updates the corresponding tests to ensure these cases are properly handled.

Handling NaN values in scoring:

• skada/metrics.py: Added checks to set ice_same and ice_diff to np.nan if their respective index lists are empty. Modified the calculation of ice_score to use np.nanmean instead of a simple average. [1] [2] [3]

Updating tests:

• skada/tests/test_scorer.py: Added a DummyEstimator class and new test cases to verify that ice_same and ice_diff are set to NaN when appropriate, and that the combined score is calculated correctly even when individual scores are NaN.