Create generic data.frame methods for sample_n/frac_keys, stratify_keys, and facet_sample / facet_strata

[njtierney]

This means that the user can specify their own keys if they like, meaning that they can do some sampling around some grouping structure of their choice.

It also means that downstream of these functions, things like facet_strata and facet_sample the user can specify a "key" argument (perhaps in ...) and this can be the grouping structure for the plots