Support cancellation requests

[MichaReiser]

Summary

This PR adds support for the LSP's cancellation requests.

This is mainly a backport of the following two ty PRs:

• [ty] Support cancellation and retry in the server #18273
• [ty] Abort process if worker thread panics #18211

Unlike ty, this PR doesn't add support for retries. Retries aren't necessary because
Ruff doesn't have to handle Salsa cancellations

• It removes Requester, Responder and Notifier and unifies them under a single Client API. I found them more confusing than helpful, and I don't see any reason why we should restrict background tasks from sending requests.
• It removes ClientSender: This is mostly a fallout of the above.
• Split IOThreads out of Connection: This removes the entire complexity around weak references because scoping rules now ensure that Connection (and its senders) are dropped before we join the thread.
• Remove the global MESSENGER and instead require callers to explicitly pass a Client. This helped find show_message call sites where MESSENGER wasn't initialized. It also removes global state, which is always good
• Responses are sent to the main loop and not directly to the client. This is required to mark the request as complete (or omit the response if the request was cancelled), and it is only possible with a mut reference to Session that background tasks don't have. This approach is the same as r-a's.
• Notifications and requests are still sent directly to the client to reduce load on the main loop. We could change that if we want but I decided against it for now.

Test Plan

Started VS code with the new ruff server.

• Verified that diagnostics still show up.
• Verified that no ruff process is running after closing VS code
• I added a 1s sleep to the diagnostic handler. I verified that there are cancellation requests (and the requests are cancelled)

[github-actions]

mypy_primer results

No ecosystem changes detected ✅

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[dhruvmanila]

Looks good! I haven't looked very closely at the code but it all looks almost same as the one in ty_server so I've provided a couple of recommendations to de-duplicate if possible. This doesn't need to happen in this PR and we can also open a "help-wanted" issue for this.

[dhruvmanila]

This looks exactly the same as the one in the ty_server, can we re-use it?

[dhruvmanila]

This also looks the same as the one in ty_server except for the retry method, maybe this could be re-used as well?

[MichaReiser]

Yes, it's almost the same with the exception that it doesn't have a retry function.

It's not a "difficult" problem but there is no good place for the files that are shared between ruff and the ty server. And Client depends on Session, which makes the reuse a bit of a challenge.