rpc: fix flaky test TestServerWebsocketReadLimit

[mishraa-G]

Fix: Tolerate "connection reset by peer" in TestServerWebsocketReadLimit

What does this PR do?

This PR updates the TestServerWebsocketReadLimit test to tolerate the "connection reset by peer" error as a valid outcome when the websocket read limit is exceeded.
This error can occur due to a race condition where the TCP connection closes before the websocket close frame is sent, as discussed in issue #32866.

Why is this needed?

Occasionally, in CI or some local environments, the test fails with:

unexpected error for read limit violation: write tcp ...: write: connection reset by peer

Previously, the test only accepted specific websocket errors (like CloseError code 1009, ErrReadLimit, or errors containing "1009"/"message too big").
By allowing "connection reset by peer", the test becomes more robust against this race condition and prevents intermittent failures.

Reference

Fixes #32866

Checklist

• Ran tests locally
• Only test code changed

---

Let me know if you want to edit or clarify anything!