More resilient reporting when revdep's DESCRIPTION is missing or empty

[jennybc]

Fixes #295, fixes #310

I had a very conservative goal here: replace the NAs we see for problem packages in cloud_summary(), revdep/README.md and revdep/failures.md with a package name. I have succeeded in this.

There are still quite a few problems with how these problem packages are being reported (or, rather, are not necessarily being reported), but I think other issues already capture some of the more substantive problems (such as #317, #301).

I was developing against a specific cloud revdep run for usethis, so I had authentic weird stuff to work with. All the NAs in my case were due to DESCRIPTION being non-existent or empty, for a revdep package. That was explained by the package being on Bioconductor, not CRAN, or requiring a higher R version than we were checking with. Or because shit happens? There was at least one package that had bizarre results for no identifiable reason and this self-resolved in a future revdep run.

[jennybc]

The GHA failure looks like a problem in r-lib/actions@v2, not anything in this PR.

[DavisVaughan]

This worked for me in tidyr! https://github.com/tidyverse/tidyr/commit/312b6d1567533f23554591cd48a7abc5ef58dcd4

An interesting note is that of the 10 NAs this filled in for me:

• 6 were Bioconductor packages
• 4 are archived CRAN packages, so I don't know why they show up here as revdeps anymore
  • https://cran.r-project.org/web/packages/evaluator/index.html
  • https://cran.r-project.org/web/packages/dynfrail/
  • https://cran.r-project.org/web/packages/expstudies/
  • https://cran.r-project.org/web/packages/mafs/index.html