Container variants with different BSgenomes

[mfansler]

A user attempting to use the r-archr container encountered issues with the fact that a BSgenome is not preinstalled in the container. Is there precedent for how to best deal with this?

Some options for how we could help with this here:

1. Naively include a bunch of BSgenomes in the run requirements so they get installed in container.
2. Build variants with different genomes.
3. Generate metapackage outputs that have different genomes in the run requirements.

I think (1) undermines users who want to run with minimal footprints. (2) and (3) would only differ by whether end-users are specifying a different package name (or container name) or specifying a build variant string.

Any thoughts about these approaches?