fix(socket/tcp): do not abort idle connections on user timeout

[Aditya-9-6]

Summary

Fixes #1202.

Per RFC 5482 (TCP_USER_TIMEOUT) and smoltcp's documented set_timeout contract, user timeout bounds how long transmitted data may remain unacknowledged, or how long an unresponsive peer is tolerated during keep-alive probing or handshake/teardown:

A socket with a timeout duration set will abort the connection if either of the following occurs:

• After a connect call, the remote endpoint does not respond within the specified duration;
• After establishing a connection, there is data in the transmit buffer and the remote endpoint exceeds the specified duration between any two packets it sends;
• After enabling keep-alive, the remote endpoint exceeds the specified duration between any two packets it sends.

RFC 5482 Section 2 states:

"The TCP user timeout controls how long transmitted data may remain unacknowledged before a connection is forcefully closed. It does not control how long an idle connection may remain open without traffic."

Root Cause

1. In poll_at(), timeout_poll_at unconditionally armed PollAt::Time(remote_last_ts + timeout) whenever remote_last_ts and timeout were present, even when the socket was in Established with an empty transmit buffer and no keep-alive.
2. In timed_out(), expiration was evaluated solely as timestamp >= remote_last_ts + timeout, causing connections to abort with RST after timeout of quiet.
3. In close(), initiating a FIN on a connection that had been idle for an extended period inherited the stale remote_last_ts, which could trigger an immediate/premature timeout before the remote peer could ACK the FIN.

Solution

• timeout_active(&self) -> bool: Added helper to determine whether the socket is in a state actively awaiting a remote peer response:
  • SynSent | SynReceived: Connection handshake active.
  • FinWait1 | Closing | LastAck: Connection teardown active (waiting for FIN ACK).
  • Established | CloseWait: Active if !self.tx_buffer.is_empty() (data in flight) OR self.keep_alive.is_some() (keepalive probing active).
  • All other states: Inactive.
• poll_at: Returns PollAt::Ingress for idle connections, and only queries PollAt::Now for timestamp initialization when timeout_active() is true.
• close(): Clears self.remote_last_ts = None when transitioning to FinWait1 or LastAck so the FIN timeout begins freshly upon FIN transmission.
• Tests:
  • Restored test_established_timeout to assert PollAt::Ingress while idle before data is queued.
  • Added test_established_idle_no_timeout verifying extended quiet periods do not abort established connections.
  • Added test_established_idle_close_timeout verifying close() after extended idle periods does not prematurely abort the FIN handshake.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.60%. Comparing base (efcba2e) to head (1f45f82).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1204      +/-   ##
==========================================
+ Coverage   82.57%   82.60%   +0.02%     
==========================================
  Files          83       83              
  Lines       26174    26215      +41     
==========================================
+ Hits        21613    21654      +41     
  Misses       4561     4561              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.