Fix CLI child process cleanup on app quit

[wojtekn]

Related issues

• Fixes flaky E2E blueprints tests (ENOTEMPTY on session cleanup)

How AI was used in this PR

Claude identified the root cause and implemented the fix.

Proposed Changes

• e2e-helpers.ts: Retry rimraf on ENOTEMPTY (up to 5 times, 500ms apart) as a safety net for any remaining race on session cleanup.

Testing Instructions

• Run the blueprints E2E tests: npm run e2e -- blueprints
• Verify no ENOTEMPTY errors appear in test cleanup
• Manually open a site, run a blueprint-based site creation, then quit the app — confirm it exits cleanly without leaving orphaned processes

Pre-merge Checklist

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

[wpmobilebot]

📊 Performance Test Results

Comparing 782188e vs trunk

app-size

Metric	trunk	782188e	Diff	Change
App Size (Mac)	1263.86 MB	1263.86 MB	+0.00 MB	⚪ 0.0%

site-editor

Metric	trunk	782188e	Diff	Change
load	1887 ms	1826 ms	61 ms	🟢 -3.2%

site-startup

Metric	trunk	782188e	Diff	Change
siteCreation	8184 ms	8161 ms	23 ms	⚪ 0.0%
siteStartup	4833 ms	4866 ms	+33 ms	⚪ 0.0%

---

Results are median values from multiple test runs.

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

[fredrikekelund]

I can't recall… The E2E tests are flaky on macOS, right? Did we try to simply pass some combination of maxRetries, backoff, maxBackoff, and retryDelay to the rimraf call in apps/studio/e2e/e2e-helpers.ts?

rimraf has a pretty generous default retry config on Windows. Much less so on macOS.

[wojtekn]

@fredrikekelund it seems it doesn't handle ENOTEMPTY:

exports.codes = new Set(['EMFILE', 'ENFILE', 'EBUSY']);

[wojtekn]

I removed changes in execute-command.ts for now to see if e2e improvement is enough to fix the issue.