Build in some sort readmapping parsing

[halexand]

The mega merge file appears has a lot of contigs that don't recruit reads in other samples. While this is probably okay at some level of the analysis, when it comes to comparing counts / coverage across samples it probably makes sense to try to work with a subset of transcripts that are relatively well covered across samples.

Maybe implement a cut that looks at initial salmon mapping and requires a transcript to have a count > 0 in X % of samples or the like?

I am also curious how this ultimately impacts read recruitment against megamerge.

[akrinos]

@halexand we've come full circle on this since 2020