Perform checks for other P value distributions

[Adam Sardar]

It would be good if the routines spat out an error when an incorrect P-value distribution was provided.

Internally this would probably look like a goodness-of-fit test between a variety of alternative models (bimodal, uniform, right skew mixture etc.).

Probably could start with the P-value oddities described here:
http://varianceexplained.org/statistics/interpreting-pvalue-histogram/