speed up analyses (particularly clustering)

[andrewheusser]

The clustering scores can take a while to compute when correcting with permutation tests. We already have a parallelized solution over analysis 'chunks' by setting parallel=True. Two ways to speed this up are:

1. optimize the multiprocessing code (may not be implemented in the most efficient way)

2. deriving an analytic solution to correct the clustering scores instead of relying on permutations