Option to downsample before fastp

[antunderwood]

In my most recent analysis, each of the fastq files for a sample was 8GB (yes I know) which led fastp to give up the ghost, I think because of memory but there wasn't an obvious error message. I'm not sure there's much we can do about this as we subsample after fastp. I guess it may be a pretty rare use case but I thought it worth mentioning