Add a couple more Feature Selection Algorithms

[abstractqqq]

Some additional feature selection algorithms of interest: Here are some ideas.

1. ROC-AUC. Run a logistic regression on one feature and use its AUC as an indicator to feature strength. This method could be generalized to feature combinations. Add support for at least 2-3 combinations of all features. Explore the runtime. We can pool the features together.
2. KS-statistics: https://www.listendata.com/2019/07/KS-Statistics-Python.html
3. Union Selector. Run a sequence of selectors and pool features together. Then use some logic (like RFE) to trim down the union set's size to desired size.