Column data should be inputs to a user-created function that transforms data and fits model

[tmastny]

This idea supercedes #5. Instead of retroactively generating documentation, leadrboard.RDS columns should act as the arguments to a transformation/modeling R script/function.

The R script will need to be identified in the leadrboard.RDS for each model, as each model could be use the same or different model templates.

[tmastny]

This originated from my (still disorganized) second Kaggle contest: https://github.com/tmastny/uno_ml_contest2/tree/master/R

I was simply saving models with only method changes to document each run.

[tmastny]

To do:

• [ ] add documentation