Implement cross_val_* methods from sklearn.model_selection

[phausamann]

These methods do not work for pipelines/estimators that change the number of samples, because of the check_consistent_length check.

[kmsquire]

I just ran into this when attempting to change the scoring function to GridSearchCV (or at least it looks like the same issue). I'll try to take a pass at implementing these.

[phausamann]

Thanks @kmsquire , a PR would be very much appreciated.