Keep CLI child close handlers alive when killing on quit

[gavande1]

Related issues

• Follow-up of Fix Windows E2E: combined status reporting, PHP INI tilde fixes, and per-command CLI shutdown #4082

How AI was used in this PR

Claude wrote the code, guided by the author. The author tested it locally and manually reviewed it before pushing.

Proposed Changes

The quit-time kill paths (appQuitHandler in the desktop and killChild in the shared runner) called child.removeAllListeners() before killing the child, which also removed the close handler. Since will-quit can be prevented and the app can keep running, this left a stale will-quit listener attached and left callers awaiting the command hanging forever. The kill paths now only kill the child and let the close handler run: it detaches the will-quit listener and emits failure, so awaiting callers settle.

Testing Instructions

• Run npm start, start a site, and quit the app. Child CLI processes should terminate as before, and the console should show a failure warning from the events subscriber instead of silence.
• npm test -- apps/studio/src/tests/index.test.ts passes.

Pre-merge Checklist

• Have you checked for TypeScript, React or other console errors?

[wpmobilebot]

📊 Performance Test Results

Comparing 5f3b4f9 vs trunk

app-size

Metric	trunk	5f3b4f9	Diff	Change
App Size (Mac)	1411.96 MB	1411.96 MB	+0.00 MB	⚪ 0.0%

site-editor

Metric	trunk	5f3b4f9	Diff	Change
load	730 ms	1088 ms	+358 ms	🔴 49.0%

site-startup

Metric	trunk	5f3b4f9	Diff	Change
siteCreation	6477 ms	6513 ms	+36 ms	⚪ 0.0%
siteStartup	2384 ms	2391 ms	+7 ms	⚪ 0.0%

---

Results are median values from multiple test runs.

Legend: 🟢 Improvement (faster) | 🔴 Regression (slower) | ⚪ No change (<50ms diff)

[fredrikekelund]

The reasoning here makes sense to me 👍 The close event listener should fall through to the else clause for cases when Studio killed the process (exitCode is null in those cases). That's good, because it means that CLI commands that don't explicitly handle SIGTERM signals will end up emitting failure events anyway, which is what we want.

[gcsecsey]

The reasoning makes sense, thanks for the detailed description. 👍