Decouple text-based data source versions from package version

[florisvdh]

Much like inbo/n2khab#183: a separate 'frictionless' data package with schemes, scheme_types and namelist will be more robust than keeping them in the R package.

Also, implementing version interdependency between scheme_types and the types data source of {n2khab} will be needed (e.g. inbo/n2khab#182 causes a different number of factor levels of the type variable in n2khab::read_types(), which has to be consistent between data sources).