Add sample weighting to V Measure Score

Describe the workflow you want to enable

Many evaluation metrics in sklearn have the ability to weight samples by some measure of importance (usually included as a sample_weight). This is also important to do when evaluating clustering. In particular, the v_measure_score (and the metrics and functions it relies on), should have this ability.

Describe your proposed solution

The proposed solution is to amend the v_measure_score function, as well as the entropy, the contingency_matrix and the mutual_info_score functions, to include a sample_weight input parameter.

This has been implemented and tested already in this repository, for this Kaggle challenge scoring metric.

Making the changes requires very little substantive changes to each function, and handling the edge cases is straightforward. This would make the v_measure_score much more useful for cases such as those in the Kaggle challenge (clustering of energy deposits of different intensities).

Describe alternatives you've considered, if relevant

A possible alternative is to keep the current functions, and handle weighting by repeating entries according to their weight. However, this is expensive and clunky, and doesn't work for fractional weights.

Additional context

No response

Start by locating the v_measure_score, entropy, contingency_matrix, and mutual_info_score entry points and their existing tests, then compare the linked implementation for sample weighting. Done means these functions accept sample_weight consistently, handle the stated edge cases, and have coverage for weighted clustering metrics.