Make it easier to ignore entire columns but keep them in returned df

[ardunn]

I might haev some column I want to ignore, such as "material_id"

I want to keep it in the df I am working with but dont want to use it for learning etc.

[ardunn]

i.e., you should be able to get back your inputs (e.g., composition) when you get your predictions back, instead of having to infer them from .loc positions