Fix ChatGPT automation executor liveness

[DevDeepakBhattarai]

Summary

• keep the ChatGPT automation executor discoverable across Manifest V3 service-worker suspension with a one-minute alarm
• wait for the support executor to reconnect after launching Chrome instead of treating process spawn as browser readiness
• extend browser presence/launch guards so a sleeping or misconfigured executor cannot trigger repeated Chrome windows
• fail with an explicit executor configuration error instead of waiting for prepare_thread to time out

Why

prepare_thread could be queued after the backend launched Chrome even when no support executor actually connected. Separately, the backend's 25-second browser-presence lease was shorter than a normal MV3 worker sleep/wake cycle. Together those paths caused intermittent prepare_thread timeouts and repeated Chrome launches despite an existing Chrome instance.

Verification

• pnpm type
• pnpm thread-sync-test
• git diff --check

The Thread Sync suite is isolated and does not start a real browser or network listener.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.