nonstandard statistical nomenclature

[benthestatistician]

I find a few nonstandard uses of experimental design terminology in the randomization distribution branch material:

• simpleRandomSampler() might be better termed "complete random sampler". This seems closer to jargon "complete block design" in agricultural experiments, as well as closer to the usage of the textbook by Gerber and Green (cf. chapter 2, Box 2.5).
• in RItest()'s type= argument, "exact" seems a misnomer: in practice the calculation is done on a sample of random assignments, not the full universe of possible assignments, so it's a simulation-based approximation. (Perhaps the answer here should involve getting rid of the type= argument: in the surprising event that someone gets around to coding up distributional approximations to go with these functions, that'll be enough of a change to warrant adding in the additional argument at that time.)
• From various places I've heard the neologism "rerandomization" to describe the sort of sampling (of assignments) being done here. I think it's evocative and useful and worth including in our docs.