sytemize the process of retraining/saving recommenders

[hjwilli]

The serialized recommenders will need to be periodically retrained- for example as we update python packages, add new experiment configurations, or update what information the serialized recs contain. We want to be able to easily rebuild the files and update github.

• [ ] Add command line option to ai.py that just trains and saves recommenders
• [ ] Add instructions/utility script that documents how to retrain the SVD recommenders for web and PennAIpy experiment configurations
• [ ] If feasible, create github or jenkins action to regenerate the files in a new branch and create a pr (possible complication is git lfs)