Run checks on a random (reproducible) subset of reads

[FranBonath]

In order to enable more compute heavy analysis methods (or alternatively just speed up the run), I think it would be very good to have a module to generate random subsets of the input read files.

I believe https://nf-co.re/modules/seqtk_sample/ can be used.

Ideally the user should be able to specify

1. A random seed to enable reproducibility
2. A number of reads to aim for per sample (if there are fewer reads in a sample, output a warning and use all reads)

[pontushojer]

Just some thoughts regarding this

Ideally the user should be able to specify

1. A random seed to enable reproducibility

I guess this is no longer relevant with PR #50 merged as the seed is currently fixed by default to 100 in the seqtk_sample module and module config, so the sampling would already be reproducible.

2. A number of reads to aim for per sample (if there are fewer reads in a sample, output a warning and use all reads)

seqtk sample will output the input without raising an error if you request more reads than are available (see lh3/seqtk#86). If we wanted to warn about this, the most sensible solution IMO would be to count the reads in the raw FASTQs separate from the seqtk sampling (e.g. using seqfu stats) and compare that to the desired sampling number to output a warning. This would also enable fractional subsampling, as suggested in #60.