Revamping postmodeling

In this issue, we are trying to consolidate the code living in disparate places into a cohesive postmodeling module and create a process for adding new functionality to this module. We need to consider Adolfo's PR #673 here as well.

Basing this on four basic questions we need to answer:
- Given two (or more) model groups, can we compare their similarities and differences? (in terms of performance, lists, errors, feature importances)
- Given two model ids, can we compare their similarities and differences?
- Given a model ID, can we describe what it's doing?