When computing the Z score in plot_pairwise_fst, if se == 0 we can get an error. It shouldn't happen on real data but it sometimes does on the simulated data leading me to believe that we should catch this error and return a 'NaN' (and that we may not simulate the data well enough).