[MLC-388] added limit on dataset features number for features importance calculation

[yromanyshyn]

short list of changes:

1)

added deepchecks.utils.features._NUMBER_OF_FEATURES_LIMIT = 100 to limit features importance calculation only for
the datasets that contain a fewer number of features than the specified limit

2)

added next functions:
+ deepchecks.utils.features.set_number_of_features_limit
+ deepchecks.utils.features.get_number_of_features_limit
to allow user to change deepchecks.utils.features._NUMBER_OF_FEATURES_LIMIT

after some simple experimentations, it become obvious that the calculate_feature_importance execution time is too dependant upon the type of the model (and probably upon of bunch of other variables) so it is not the best idea to leave the user without the possibility to redefine that limit.

collected execution time statistics

3)

after the above changes, some modifications were also needed for the next checks
- WholeDatasetDrift
- SegmentPerformance

4)

refactored calculate_feature_importance function

[CLAassistant]

All committers have signed the CLA.

[noamzbr]

approved pending setting the default to 200