Stop all threads and write ANSI-Reset on Ctrl-C

[th1000s]

On Ctrt-C stop all threads currently writing to the terminal, ignoring whatever they have locked, and write the ANSI-Reset code directly to the terminal, then exit.

This uses pthread_kill() or SuspendThread(), then libc::write() or WriteConsoleA() on Unix or Windows, respectively.

---

So far only tested and enabled on Linux, macOS and Windows (cmd.exe, Powershell, Mingw64), but I am sure this will work on the BSDs and several others which have pthread_kill() - but I'd rather have someone confirm this first.

And to compare the before/after, running rg with a -j of 7, 11, 22, 33.. etc. will disable this handler.

[th1000s]

When suggesting to someone to use rg instead of "legacy" grep -R .. | grep -v -e .git -e build or git grep .. on the very first try (maybe spooked by the speed of the output) after ^C the terminal remained blood red and I had a slightly embarrassed look on my face.

A lot of users have shells or prompts which handle this automatically. But now that rg is packaged by most Linux distros, more users which don't customize their shells use it, and first impressions matter as to not end up as "newfangled nonsense" :)

[BurntSushi]

I've only skimmed this PR so far, but there is a ton of code here. And it also introduces a new dependency on winapi, which I hope to be getting rid of soon.

Separate from that, ripgrep used to have handling like this, but it ended up offering a worse user experience. Namely, ^C wouldn't always terminate ripgrep. And I believe it worked very poorly on Windows.

A lot of users have shells or prompts which handle this automatically.

I just use a simple little script to fix coloring if and when it happens (which is rare): https://github.com/BurntSushi/dotfiles/blob/eb14900839a210f31b826f0ab070506528ea6a99/bin/reset-term

But now that rg is packaged by most Linux distros

This is not a recent development. This has been true for years.

[th1000s]

winapi -> windows-sys

Both crates work fine, I added a commit which uses windows-sys instead.

Separate from that, ripgrep used to have handling like this, but it ended up offering a worse user experience. Namely, ^C wouldn't always terminate ripgrep. And I believe it worked very poorly on Windows.

I've seen the previous behavior, this is fixed in this PR. It does this by (safely) ignoring any held locks which could slow down a fast exit, so a few syscalls have to be issued directly.

[rg packaged by most Linux distros is] not a recent development. This has been true for years.

I must be thinking in Debian releases, there it has only been included for one or two releases I think :)

[BurntSushi]

ripgrep has been in Debian for years.

I'm not sure if I'm going to take this. It is so much code to solve an issue that I don't think is a big deal.