Update to R 4.2 & Bioconductor 3.15

[jashapiro]

We may want to update the R version in the scpca-renv conda environment to 4.2, which will force an update to Bioconductor 3.15. The reason for this would be to allow compatibility with anndata 0.8.0, which is incompatible with the version of zellkonverter in Bioconductor 3.14, as discussed below.

With this update we would also want to update other environments to anndata 0.8.0

Doing this will probably mean that we will also want to update the R version & packages in scpca-downstream-analysis for consistency, but that would have to happen independently.

Starting my review, and want to note this first. According to https://github.com/theislab/zellkonverter/releases, the update to anndata 0.8.0 is a breaking change. To use it with zellkonverter, we would need to update to R 4.2 & Bioconductor 3.15. I would actually think this is probably where we want to be for future compatibility, and it is probably easier to make this change now than later. Better not to get stuck too many versions behind. If others agree, then I can file an issue to update the environments (stacked on #64, I expect, unless that gets in first), and we can keep this environment (and scVI, presumably) at anndata 0.8.0

Ah good find! I'm okay with this. I am slightly worried about developing locally for scpca-downstream-analyses, which requires R 4.1.2, and then having to develop here with R 4.2. Unless that shouldn't be a problem? I just know that when you use renv::snapshot() it stores the R version in the lockfile which could cause adding things to the lockfile in scpca-downstream-analyses to be more difficult, unless you know of another trick? Or should we update both repos...?

Originally posted by @allyhawkins in #72 (comment)

[jashapiro]

Looks like this will have to wait a bit; the R 4.2 version on condaforge has been held up by windows incompatibility, but it looks like that will be released soon: conda-forge/r-base-feedstock#205

[jashapiro]

Completed in #92