wrapping multicam procedure around other regression models?

[ismael-mendoza]

As part of #153 , we realized that wrapping the MultiCAM pre-processing and post-processing procedure around a ML model (in this case xgboost) wasn't as trivial as we once thought. Perhaps we could explore this further and understand why this doesn't simply improve performance.