#248 has brought to my attention that the types, as is, are seemingly equivalent to classNames(...args: any[]): string.

It'd be nice to have some tests to verify that isn't the case, otherwise, we could probably simplify the definitions...