Add pseudo_count to log2_FC so zero mean genes don't break fold changes

[mertcan-demir]

Hi! First time contributing, so please let me know if I missed anything. I noticed log2_FC() throws Inf when one group has zero mean expression. So I added a pseudo_count parameter as a quick fix. The default is 0 so nothing changes. Not sure if this is the best way to handle it (maybe preprocessing the assay is cleaner like proportional constants as edgeR do with prior.count?), but it solved the issue on my end without touching the mean_* columns. Happy to adjust if you prefer a different approach! Changes are only in R/log2_FC.R one new param + validation + a one-liner in the FC calculation.

[SimoneTiberi]

Hi Mert, thanks a lot for spotting this issue and informing us!
I appreciate your contribution: I'll look at it in June and implement a solution (yours or something similar) both in our Github and Bioconductor repositories.