test(mcp): cover --sandbox at the config resolution level

[Yury Semikhatsky (yury-s)]

Summary

• --sandbox enables the sandbox has been failing on ubuntu-latest - chrome since it landed: it starts an MCP server with --browser=chromium --sandbox, and the server launches the browser eagerly during initialize. Chrome for Testing cannot start sandboxed on Linux (No usable sandbox!), so browser_get_config returned an error instead of a config.
• Replaced it with tests in config-resolve.spec.ts that drive real argv through decorateMCPCommand into resolveCLIConfigForMCP, covering --sandbox, --no-sandbox and neither flag without launching a browser.

Failing ubuntu-latest - chrome jobs:

• https://github.com/microsoft/playwright/actions/runs/32402013434/job/96532183849
• https://github.com/microsoft/playwright/actions/runs/32396905156/job/96515743225
• https://github.com/microsoft/playwright/actions/runs/32395156145/job/96510147300
• https://github.com/microsoft/playwright/actions/runs/32371792173/job/96433836023
• https://github.com/microsoft/playwright/actions/runs/32321251720/job/96283729180

[github-actions]

Test results for "MCP"

1 failed
❌ [firefox] › mcp/cli-core.spec.ts:70 › dblclick @mcp-windows-latest-firefox

8128 passed, 1361 skipped

---

Merge workflow run.

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failure here.

🟢 The one failure is a pre-existing flake — this PR is clear

[firefox] › mcp/cli-core.spec.ts:70 › dblclick (on mcp-windows-latest-firefox) is unrelated to this PR. This change only touches MCP sandbox config-resolution tests (config-resolve.spec.ts, config.spec.ts), and dblclick exercises none of that. The same test fails intermittently on SHAs this PR can't be responsible for.

Details

Pre-existing flake / infra

• [firefox] › mcp/cli-core.spec.ts:70 › dblclick — cross-run flake on mcp-windows-latest-firefox: failed 3 of 614 runs (~0.5%), passed the rest. Two of those failures are on unrelated commits — PR feat(trace): add action screenshot and aria snapshot events #42167 (sha e98b16d7, run 31228415523) and a main push (sha c8fc3bf8, run 31559512534) — with the identical signature (expect(received).toContain(expected) … Received has value: undefined, sometimes with a 60s timeout). The third is this PR's run (32407839643). Same symptom, unrelated SHAs → this is the test flaking, not your change.
• This PR edits only sandbox config-resolution tests and launches no browser in them, so it can't reach the cli-core double-click flow.

Recent history on this bot (oldest → newest, this PR's run is the trailing 🟥):

🟩🟩🟩🟩🟥

_{Triaged by the Playwright bot - agent run}