Fix Normalized enrichments scores and repo improvements

[iblacksand]

This release fixes a bug for cases where the GSEA enrichment score for a gene set did not share a sign with any of the random permutations.

Previous behavior

Example case: the enrichment score for set $S$ is a negative value. The 1000 permutations had all positive enrichment scores. To calculate the normalized enrichment score (NES), you divide the real score by the average of the random scores that are of the same sign (in this example the average of the negative scores). Since there are no negative random scores, the average would be 0.

To avoid dividing by 0, old WebGestaltR would add 0.000001 to every average. In this example, this would result in multiply the enrichment score by $10^{-5}$.

The old R package would set the NES to 0 if this happened, but this step was missed in the rust implementation, resulting in a high NES.

New behavior

To avoid this abnormally high normalized score, if there are no permutations with a matched sign, the normalized score is set to 0. This is the same behavior as the original R package.