Improve versioning of models - save pre-processing code along with models

[helloxinw]

We save many versions of our ML model. However, the cleaning and pre-processing code isn't saved. This makes it hard to use old models and re-create the result, especially now that different features are being tested.

Also, add the Python version to pickle file names, or maybe list it in some documentation. We need the correct Python version to load the pickle.

Solution
Perhaps a pickle file or something could be used to save the state of the code.