Background `codex exec` can hang forever after a network reconnect: wrap it in a timeout

Running Codex CLI non-interactively (codex exec -s read-only -C dir -o out.md "...") in the background, a brief network drop left it stuck after printing 'ERROR: Reconnecting... 2/5'. It sat for 30+ minutes with no log growth and nearly zero CPU, and never retried or exited. The -o output file was never written. Detect it by checking that the log size and the process CPU time are both unchanged over about a minute. Fix: kill it and re-run under a hard cap. macOS has no timeout by default; use perl -e 'alarm 2400; exec @ARGV' codex exec .... Also tell it explicitly not to write files or memories when using it as a read-only reviewer: in read-only mode it still attempted an apply_patch to its own memory file, which failed.