Handle KaTeX warnings

[bep]

Fix #13735

[jmooring]

This is great. My only comment is that you can:

1. Uncomment this:
   https://github.com/gohugoio/hugo/pull/13760/files#diff-8bb30d10edead28c083227997d4d8788e9d0c82235ad17ff85dd7e14a0d3132cR529-R532

2. Remove this:
   https://github.com/gohugoio/hugo/pull/13760/files#diff-8bb30d10edead28c083227997d4d8788e9d0c82235ad17ff85dd7e14a0d3132cR528

3. Remove this:
   https://github.com/gohugoio/hugo/pull/13760/files#diff-8bb30d10edead28c083227997d4d8788e9d0c82235ad17ff85dd7e14a0d3132cR539-R556

[jmooring]

Please hold off on merging this for a bit. I found something related to the (unexpected) way that KaTeX handles the newLineInDisplayMode error.

Note to self, again: disable cache.misc while testing.

See https://discourse.gohugo.io/t/newline-problem-with-transform-tomath/54833 for newLineInDisplayMode error example.

[bep]

@jmooring Feel free to push to this PR (or if needed, just take my branch and create a new PR).

[jmooring]

When strict is set to warn, the warning appears the first time you build, but once the result is cached the warning does not appear again. When we retrieve the cached result in future builds, we have no idea if warnings were emitted during the initial run.

This is certainly solvable (e.g., caching a struct in a gob, caching JSON, etc.), but I'm not sure it's worth the effort.

Unless there's a low-effort way to handle this, I think limiting the strict options to error (the default) and ignore is good enough.

[bep]

@jmooring have a look at my last commit.

[jmooring]

This works really well. Ready to squash and merge when tests go green. Thanks.

[bep]

@jmooring I didn't know that the placment of the "co author" line mattered when squashing mattered, but now I know.

[jmooring]

@bep On the list of things that concern me, this is number 4374.