add an affirmation for (lack of) variation

[tgerke]

This check is suggested in ICH E6
p. 32 "monitoring of clinical trial data" (probably some other good ideas to implement here too)

[ddsjoberg]

That's a good one. If you outline the API, I can implement.

1. Numeric variables could be something like sd(x) >= t for some threshold, t.
2. Character variables could be length(unique(x)) >= t