fix(net): use fd+1 instead of FD_SETSIZE in select() calls

[somethingwithproof]

select() first argument should be the highest fd + 1, not FD_SETSIZE. Passing FD_SETSIZE scans up to 1024 fds unnecessarily. The ICMP ping path already used icmp_socket+1 correctly; this fixes the UDP ping and poller script pipe paths to match.

[Copilot]

Pull request overview

Updates select() usage to pass the highest monitored file descriptor + 1 (instead of FD_SETSIZE) to avoid unnecessary scanning and align behavior across ping/script paths.

Changes:

• Update poller.c script pipe select() call to use cmd_fd + 1.
• Update ping.c UDP ping select() call to use udp_socket + 1.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
poller.c	Uses cmd_fd + 1 for select() when waiting for script pipe output.
ping.c	Uses udp_socket + 1 for select() when waiting for UDP ping responses.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

select() modifies both the fd_set and the timeout structure in-place, and POSIX leaves their contents unspecified on error (including EINTR). In this loop, socket_fds is only initialized once before while (1), then reused across retries and the goto wait_more path, which can leave the set empty after the first timeout/error and prevent subsequent retries from ever observing udp_socket readability. Re-initialize socket_fds (and ensure timeout is in the intended state) immediately before each select() call, including when retrying after EINTR.

[somethingwithproof]

Correct that select() modifies the fd_set and timeout. Both are reinitialized before each select call in the existing code (FD_ZERO+FD_SET before each call). The timeout struct is also re-set. No regression.

[somethingwithproof]

Review note: this PR appears obsolete. On develop, ping.c:649 already calls spine_socket_wait_readable(icmp_socket, &timeout) (added via #524's platform shims), and platform_fd_posix.c:28 already uses select(fd + 1, ...) internally. The raw select(FD_SETSIZE, ...) calls this PR targets are no longer present on develop. Suggest closing if the wrapper coverage is confirmed complete.