Wrapper for scikit-learn models

[nlsschim]

Wrapper module that allow a user to choose any model architecture listed on scikit-learn, and train it on MPT data. Specifically, this should ensure that, regardless of what model is used, that hyperparameter tuning and train test validation splitting approperiately ensure no leakage of trajectories that are in the same microenvironment. This code exists in the repository for XGBoost, but should be extended.

Another use case for this code would be to enable a user to easily try out an ensemble of models to see what performs best (ie random forest, Recurrent neural net, simple feed forward neural network). Code for Random Forest and RNN should exists from a past students work (David), and code for a convolutional neural network also exists from code Nels wrote.

[nlsschim]

I think this package could have some useful ideas we can use!