Create a test for #11264

[jdm]

It's not clear to me why it's difficult to test. We had a reproducible test case, and the code changes seem to be working around a particular situation that can occur. Why can't we reverse-engineer the conditions that make that code be executed?

cc @KiChjang

[KiChjang]

I'm really not sure why the logs in #11261 show that there was a .borrow() call from update_href before. I double-scanned and triple-scanned htmlanchorelement.rs before #11264 was merged, and I still don't see any borrow() at all in update_href.

I wasn't able to reproduce it on my computer since I'm using Linux and I hit a FontCache panic.

[jdm]

I have not been able to reproduce #11261 yet from 904e9ea. The panic backtrace reported by @paulrouget doesn't make sense at all, any @KiChjang wasn't able to reproduce the failure so the fix was produced on faith. I would like to properly figure out the actual problem we solved; @paulrouget can you revert to the revision in which you observed the panic and get a proper backtrace from rust_panic on a debug build?

[nox]

Was this done?

[jdm]

It was not done.

[jdm]

This will never be done.