Add a bunch more bibtex tests

[CraftSpider]

This includes tests for crossrefs, multiple cites, and some bad aux lines. This should cover most of the important main code paths as well as some particularly useful warning/error paths.

[CraftSpider]

These tests aren't just for my bibtex rewriting work - this batch catches another real error due to cite_xptr never getting set to 0 on init, which I suspected when I noticed it in rewrite, but running these tests confirms it, as it causes run-order dependent failures.

[codecov]

Codecov Report

Merging #1039 (84604ac) into master (b8b0245) will increase coverage by 0.25%.
The diff coverage is n/a.

❗ Current head 84604ac differs from pull request most recent head acb9744. Consider uploading reports for the commit acb9744 to get more accurate results

@@            Coverage Diff             @@
##           master    #1039      +/-   ##
==========================================
+ Coverage   44.84%   45.09%   +0.25%     
==========================================
  Files         155      155              
  Lines       62605    62609       +4     
==========================================
+ Hits        28074    28234     +160     
+ Misses      34531    34375     -156     

Impacted Files	Coverage Δ
crates/engine_bibtex/bibtex/bibtex.c	50.67% <ø> (+5.91%)	⬆️

... and 4 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[CraftSpider]

Error appears to be that on Linux, whatever opening a directory returns, it doesn't cause issues - but on Windows, you get PermissionDenied, which means it returns an error instead of trying with the extension. I'm just gonna move that test out of the directory for now, since it's a one-off anyways.

[pkgw]

Great stuff! It looks like the main CI tests passed for the most recent iteration, but GitHub is only showing me the ArXiv check. I'm going to close this and then reopen to retrigger the CI just for completeness.

[CraftSpider]

Yeah, from what I could tell, I pushed twice too close and the tests ran on the second commit but got attached to the first. Either way, looks happy now.