speed up exact p-value for small n in Kendall

[Peter]

Class Kendall currently uses a naive algorithm that parses through all N! permutations and calculates the score. Since the score calculation scales (N², see #710) this becomes expensive for not very large N. It should be possible to calculate the P-value in an recursive fashion similar to ROC. ​statsdirect tabulates values for small N and then use an Egenworth expansion for larger N (approximation?).