Normalizer parameters require a manual cast

A trained normalizer does not have public methods or properties that allows access to the normalization parameters. For example, a minmax normalizer doesn't have the min, max, or zero-level available.

I would expect that the trained transformer would have a `.Model` property, similar to a trained learner.

This is required V1 Scenario (#2498):
- I can inspect the normalization coefficients of a normalizer in my pipeline without much work. Easy to find via auto-complete.

@shauheen @eerhardt @TomFinley I marked this as P13, but it could be a V1.1 as it does not change APIs, but needs new APIs. This is similar to LDA (#2197), and I think we have a more global problem where trained transforms don't expose a `.Model` property similar to our learners.