Add a vignette to explain how to compute the number of possible distinct values of the test statistic.

[astamm]

Following Phispon & Snyth 2010, Section 6, there are some common situations in which care should be taken when computing the number of possible distinct values of the test statistic. It could be nice to write an article about that.

[astamm]

Also, should we set a different number of possible distinct values for each test statistic in the NPC case? In this case, care must be taken in still using the same permutations for each test statistic. So maybe we should allow NPC only when permutation are drawn randomly.