server: add a global request timeout

[svix-jbrown]

helps provide backpressure if things are going really south

[github-actions]

	Tests	Passed ✅	Skipped	Failed	Time ⏱
Diom CI JUnit	307 ran	307 passed	0 skipped	0 failed	5m 18s 902ms

Test	Result	Time ⏱
No test annotations available

[github-actions]

Test Coverage: 79.95%

Artifact download

[svix-jplatte]

This PR will conflict with my next one for error types. If CI wasn't failing, I'd probably just approve and merge and deal with it myself. Since it isn't, maybe it makes sense to wait for my PR and rebase this one after? It will change things a fair bit, but shouldn't be too annoying.

[svix-jplatte]

The error refactor work I mentioned all landed. If you could update this to have fn request_timeout delegate to server_error, just like fn not_ready and fn shutting_down, that would make this fit right in.

[svix-jbrown]

The error refactor work I mentioned all landed. If you could update this to have fn request_timeout delegate to server_error, just like fn not_ready and fn shutting_down, that would make this fit right in.

Sorry — I realized that because we support long-polling, I need to make this more complicated to exempt the long-polling endpoints. I'll mark as draft until I come up with an ergonomic way to do this