ShapRFECV adds support for HalvingGridSearchCV

[Matgrb]

Problem Description
With the new release of sklearn there is a new HalvingGridSearchCV. We want probatus ShapRFECV to support use of it

Solution Outline
In order not to have to bump the version of sklearn we can check if the clf is instance of BaseSearchCV, which HalvingGridSearchCV extends. In case yes, then it should be treated the same way as RandomizedSearchCV.
One also needs to update the docs.

[timvink]

HalvingGridSearchCV should be a drop-in replacement for RandomizedGridSearch or GridSearch already, according to the sklearn docs. Good to add some tests to probatus for this, at the very least.

That said, it might be easier to implement the ShapRFECV using the new SequentialFeatureSelector to make the sklearn compabitility sklearn's problem. But then we would probably have to drop some features like step and the hpopt at each step, and change the API.

[Matgrb]

Currently the HalvingGridSearchCV is not supported unfortunately, this needs to be added.

The SequentialFeatureSelector works in a different way than ShapRFECV and has much higher computational requirement

ShapRFECV for each feature set, you fit a model, find feature importance and remove least important feature. Feature importance is a proxy for how strongly the feature affects the performance of the model.

In SequentialFeatureSelector (backwards) start with the full feature set and check how removing of each feature would affect the performance of the model (using CV). This means that at each iteration, you have to remove each feature
separately, fit a model and evaluate it. This is much more computationally intense.

[Matgrb]

That issue can be solved by solution proposed in #76