Full-significance testing

[dglmoore]

Right now, we compute significance using a permutation test with only 1000 permutations. This results in an uncertainty of about ±1% at p=0.5. We need to allow the user to specify the number of permutations to use, so they can tune the uncertainty. Further, this is going to require the use of a worker pool in order to keep the UI from lagging.