Make checking GitHub linguist results for onboarded packages an editor task

[maelle]

If a package isn't classified as R repo, then tell the author to make a PR to https://github.com/github/linguist

[maelle]

hmm well not necessarily a PR cf their troubleshooting guide

[maelle]

Now I even wonder if some of the overrides should always be used for R packages (for man, and docs/ when using pkgdown) 🤔

[maelle]

Besides, it could be an author task I guess.

[maelle]

More specific guidance could be:

• The repo might be classified as C/C++ repo when the author(s) wrote a lot of C/C++ code which is fine I guess.

• When the repo is classified as say JavaScript because there's e.g. "vendored" JavaScript code i.e. not written by the author use https://github.com/github/linguist#generated-code

• When the repo is classified as html because of html documentation files that are not in docs/ (docs/ is viewed as documentation by GitHub linguist by default) use https://github.com/github/linguist#documentation

[maelle]

The guide should mention that GitHub language classification for a repo is binary. That said there are more specific language statistics accessible via clicking on the languages colored line on the repo page, and via the API(s).

Besides we put the "r" and "rstats" topic on all the package repos in ropensci so that even if repos are not discoverable as R repos via the language they are via the topics.

This is all likely not mission critical since few people browse/search GitHub but it's easy to mark repos this way hence these guidelines.

[maelle]

Added a GitHub grooming section to the guide, and added to the editor's guide in the book that they should point authors to that section of the book.