test runner: cleanup on Ctrl+C/SIGINT

[laniakea64]

Attempt to fix #45

I took another look into this, and this time something clicked about how to use ctrlc crate (which BTW just also uses), so decided to give this a try. Still don't really know what I'm doing in Rust, so @NoahTheDuke your code review would be much welcome, thanks 🙂

About test runner performance and this PR:

• I have no idea why Vim's stderr being piped is such a performance improvement over null?? 👀 This doesn't seem to be the case for stdin or stdout.
• Performance is more varied with this PR. Most runs are about the same performance as before, but I'm seeing variation up to ~0.03 seconds difference either way, with the faster side occurring more often than the slower side. I wasn't able to completely avoid having any "notably" (as in meaningful timing measurement) slower runs.
• Not sure to what extent the optimizations ended up being tuned for my specific dev environment rather than more generally? 😨
• Cargo deps were not updated again as part of this PR, because doing so this time actually reduced performance. There is a cargo-audit warning that we are using a yanked crate (ahash version 0.8.3, reasoning for the yank explained on their wiki), so the question of updating Cargo deps again should be investigated, but let's address that separately.

[NoahTheDuke]

these changes look good to me!