[roll] Roll fuchsia [netstack][fdio] TCP connection reset handling fixes

Fix fdio and netstack2 layers to propagate reset error state and return
appropriate errors for socket read/write calls, while keeping the
behavior in parity with Linux. Address few edge cases for connecting
vs connected states and blocking vs non-blocking socket calls.

- Define new fdio states to record reset error state for connected
  and connecting sockets. Apply that for subsequent read/write socket
  calls to return appropriate errors. This change ensures that the
  behavior is in parity with Linux.
- Fix netstack2 to account for connection handshake failures while
  signalling an incoming event to the client.
- Define 2 new signals to convey connection reset and refused cases
  to the fdio layer.
- Refactor part of gVisor endpoint error handling from within loopXX
  routines and offload the actions to the EventHUp callback that is
  currently registered with gVisor stack.
- Fix EventHUp callback registration to check for prior endpoint
  error state and if so, intitiate closing of the endpoint.
- Enable TCP shutdown for a listening socket.

Original-Bug: 35594
Test: fx test -od netstack-bsdsocket-c-api-test -- --gtest_filter=NetStreamTest/StopListenWhileConnect.*
Original-Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/437660
Original-Revision: c99fe9c1284265fe5eb0cd3ce7c3ba431b2f3f4a
GitOrigin-RevId: 4558fbf61ddc0e142baebcdc205223597e52f6f8
Change-Id: Ifdcbc698d02cfc03530892e6294d3cdba4bcdd12