Is your feature request related to a problem? Please describe.

It could be nice to warn users that a metric has a high variance in specific split of a cross validation.

Describe the solution you'd like

Sample detection code by @glemaitre.

def detect_outliers_mad(scores, threshold=3.5):
    median = np.median(scores)
    mad = np.median(np.abs(scores - median))
    modified_z_scores = 0.6745 * (scores - median) / mad

    outliers = np.where(np.abs(modified_z_scores) > threshold)[0]
    return outliers, modified_z_scores

This new metric should also be serialized by CrossValidationReportPayload.

Describe alternatives you've considered, if relevant

No response

Additional context

No response