Add invariants for detecting invalid combinations of columns with column groups

[pradeepnschrodinger]

Hmm, I haven't heard too many complaints / confusion around this so I don't think it's necessary to add to the old setup, although it is nice to have.

Originally posted by @wcjordan in https://github.com/schrodinger/fixed-data-table-2/diffs