ENH: Setting a model estimator to a Pipeline should decompose it if it follows our structure

[andersbogsnes]

Is your feature request related to a problem? Please describe.
When creating new model objects inside ml-tooling, we break the feature_pipeline, estimator separation set up by the user. Any time ML-tooling creates a new Model internally, it should also set the estimator and feature_pipeline separately

Describe the solution you'd like
When Model.estimator is assigned to, if the value is a Pipeline composed of "features" and "estimator" steps. it should split them up, and set feature_pipeline and _estimator.

Alternatively, all ML-tooling functionality should be aware of the feature_pipeline/estimator split and pass them around