GMEA is too slow

[markziemann]

The current way the test runs is very slow. I think the code could be substantially optimised:

1. Optimise how the t-stat vals are subset for each gene.
2. Optimise the wilcox test by re-writing the function to remove unnecessary stuff.
3. If that is still too slow, we can use C helper code to speed up parts of the process (as done by FGSEA https://www.biorxiv.org/content/10.1101/060012v3)