Love the idea!

I would ask for a little more separation of concerns. One package for raw but cleaned data with a collection of schemas, and a second for loading arbitrary data + schemas in to Django (and probably accomplishing all of the extra administrative steps provided in the example).

That way if I want to add other schemas for a non-Django use in the same package (say if I care more about analysis than clicky-interfaces) or not use Django, I can still use a package manager for the same data.