Build test code that leverages coderdata to evaluate linear model prediction of drug response

[sgosline]

I'd recommend starting with this code base: https://github.com/PNNL-CompBio/BeatAMLproteomics/blob/main/pybeataml/run_models.ipynb

And extending to work with coderdata. then we can evaluate:
1- cancer blind splits
2- effects of balanced splits
3- ability to predict organoid response

then we can compare with the deep learning models.

[sgosline]

This is filed as a documentation bug for now. We have a notebook for the linear models but need more summary statistics to really showcase the predictive capability across all drugs.