Implement two-phase graceful HTTP/2 shutdown

[arturobernalg]

This change implements two-phase graceful shutdown for HTTP/2 connections.
The multiplexer sends an initial GOAWAY, waits one RTT using PING, then sends the final GOAWAY with the last processed stream identifier while continuing to handle in-flight peer streams and peer GOAWAY boundaries correctly.

[ok2c]

@arturobernalg Why is this needed? What does this improve excactly?

[arturobernalg]

Why is this needed? What does this improve excactly?

@ok2c
RFC 9113 recommends two-phase graceful shutdown for HTTP/2: initial GOAWAY with 2^31-1, wait one RTT, then final GOAWAY with the actual last processed stream id. This improves correctness of connection draining by avoiding shutdown races around in-flight stream creation and by preserving an accurate final GOAWAY boundary.

https://www.rfc-editor.org/rfc/rfc9113.html#section-6.8

“A server that is attempting to gracefully shut down a connection SHOULD send an initial GOAWAY frame with the last stream identifier set to 2^31-1 and a NO_ERROR code.”

“After allowing time for any in-flight stream creation (at least one round-trip time), the server MAY send another GOAWAY frame with an updated last stream identifier.”

[ok2c]

A server that is attempting

@arturobernalg Very well. A server. And what about clients? Does this make any sense for client connections?