Add default tests for all combinations of add_p.tbl_difference(group, adj.vars)

[ddsjoberg]

Or at least add more defaults. I think we could probably updated ancova to use by variables to more than 2 levels, but still need to be careful since add_difference() still requires 2 levels. If we expand to more than 2, the by variable will need to be made into a factor to get the proper p-value