TestServerWebsocketReadLimit failure due to race condition

[hero5512]

System information

Geth version: master branch
OS & Version: Any
Commit hash: current master branch

Expected behaviour

The test TestServerWebsocketReadLimit should pass consistently in CI and local environments.

Actual behaviour

The test fails intermittently in CI with:

--- FAIL: TestServerWebsocketReadLimit (0.01s)
    --- FAIL: TestServerWebsocketReadLimit/limit_with_large_request_-_should_fail (0.00s)
        server_test.go:278: unexpected error for read limit violation: write tcp 127.0.0.1:33636->127.0.0.1:39835: write: connection reset by peer
FAIL

The test expects specific error types (CloseError with code 1009, ErrReadLimit, or errors containing "1009"/"message too big") but fails when it receives a "connection reset by peer" error.

Steps to reproduce the behaviour

CI failure example: https://github.com/ethereum/go-ethereum/actions/runs/18361445866/job/52306116362?pr=32856

Root Cause Analysis

The error may be caused by the TCP connection closing before the WebSocket close frame is sent, resulting in a “connection reset” instead of a proper close message.

Proposed Solution

Update the test to accept connection reset error

[mishraa-G]

Hi all,
I've opened PR #32889 to address the intermittent failures in TestServerWebsocketReadLimit as discussed in this issue.
This PR updates the test to also accept "connection reset by peer" as a valid error outcome when the websocket read limit is exceeded, making the test more robust against race conditions observed in CI.
Feedback and review are welcome!