train XG boost tree on XBT data

[stevehadd]

XG boost represents the current state of the art in decision tree ensembles. It has a scikit-learn like interface, so should be easy to integrate into the current workflow. tutorial: https://machinelearningmastery.com/develop-first-xgboost-model-python-scikit-learn/

[stevehadd]

we will need to do some hyperparameter tuning to get the best results: https://www.kaggle.com/prashant111/a-guide-on-xgboost-hyperparameters-tuning https://blog.cambridgespark.com/hyperparameter-tuning-in-xgboost-4ff9100a3b2f

[stevehadd]

With the refactoring required to accomodate gradient boosted trees, we should rerun all the classification notebooks and experiment definitions to check that they still work. The refactoring also allows us to use the scikit-learn gradient boosted trees implementation.