Updated EDA with first run feature drops

This initial-pass EDA is complete.  It created a train_prep dataset with less features (Final count 95 from 186). The next notebook will use this dataset to do prep:  Feature engineering (bin hospital_id per Kristin's code), Normalization and Smote or ADASYN to address class imbalance.