Adjust permutation p-values for bias caused by permutations resulting in 0

[mlotinga]

Permutation p-values sometimes result in 0, due to the finite number of permutations. This should not be the case, according to a detailed account of the problem as detailed by Phipson and Smyth (2010, http://doi.org/10.2202/1544-6115.1585)

Phipson and Smyth describe solutions to this problem, and provide corresponding R code (https://rdrr.io/cran/statmod/src/R/permp.R). I've implemented their approximate integral solution for effect size objects (4067f67#diff-07c1c6b6ceedbcec253b34a827f0c8ce921361387e478c3eb51986d1291e6080R1482)