Tell a more interesting story with the deep learning workflow

[wlandau]

The model tuning in the current deep learning chapter does not noticeably affect model performance. If we found more impactful things to tune, we could tell more of a story. Related: https://github.com/leonjessen/RPharma2019/blob/master/Exercises/04_diamonds_regression.md

[wlandau]

On reflection, it does not seem worth going back and redoing the customer churn chapter for this. Could be a good blog post though.