fix(test-runner): run teardown projects when maxFailures is reached

[yury-s]

Summary

• Teardown projects were skipped when a run was interrupted by --max-failures, so their cleanup (e.g. clearing a database) never ran.
• Teardown project test groups are now exempt from the max-failures interruption (via a TestGroup.ignoreMaxFailures flag), so they still run while everything else stops.

Fixes #41713

[github-actions]

Test results for "MCP"

1 failed
❌ [chrome] › mcp/annotate.spec.ts:110 › should abort annotation when last screenshot is removed @mcp-macos-latest-chrome

7759 passed, 1249 skipped

---

Merge workflow run.

[github-actions]

🟢 CI is clear — the one failure is a known flake

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

The only red on the latest commit (414dab2) is one MCP test, and it's a pre-existing flake unrelated to this PR. tests 1 was still running when I looked, but nothing so far points back to the diff. This PR only touches the test runner's max-failures / teardown scheduling (dispatcher.ts, tasks.ts), which the MCP annotation flow doesn't exercise.

Details

Overall: no failure looks caused by this PR. The single failing test is bimodal across the results DB and fails on many unrelated PRs and push builds.

Pre-existing flake / infra

• [chrome] › mcp/annotate.spec.ts:110 › should abort annotation when last screenshot is removed (@mcp-macos-latest-chrome) — flaky. Across the aggregated CI results this test flips verdict on every project without this PR's changes: chrome 5 failed / 214 passed (219 runs), firefox 10/204, chromium 2/216, webkit 2/218. The failures land on SHAs the PR can't be responsible for — e.g. unrelated PRs feat(webkit): roll to r2327 #41684, fix(screenshot): fix webp transparency and lossless detection, add tests #41781, feat(html): make merge files a reporter option instead of a UI setting #41799 (chrome) and devops: report CI test results as parquet artifacts #41669, chore(recorder): mint page aliases in the language generators #41728, fix(transform): detect partially written compilation cache entries #41739, test: roll stable-test-runner to 1.62.0-alpha-2026-07-13 #41753 (firefox/webkit), plus several push builds. On the failing chrome bot alone it's 5 red among 216 runs, the reds scattered through the whole window; the last one is just this PR's run. This PR changes teardown-project scheduling under --max-failures in the test runner, which the MCP annotate/screenshot flow doesn't touch.

Nothing to fix here on the PR's side.

Exit, pursued by a bug. 🎭

_{Triaged by the Playwright bot - agent run}

[github-actions]

Test results for "tests 1"

6 flaky

⚠️ [chromium-library] › library/video.spec.ts:294 › screencast › should capture navigation `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/chromium/oopif.spec.ts:282 › should click `@realtime-time-library-chromium-linux`
⚠️ [chromium-library] › library/video.spec.ts:356 › screencast › should work for popups `@chromium-ubuntu-22.04-node24`
⚠️ [chromium-library] › library/video.spec.ts:699 › screencast › should capture full viewport on hidpi `@chromium-ubuntu-22.04-node24`
⚠️ [chromium-library] › library/video.spec.ts:664 › screencast › should capture full viewport `@chromium-ubuntu-22.04-node20`
⚠️ [playwright-test] › ui-mode-trace.spec.ts:827 › should update state on subsequent run `@ubuntu-latest-node22`

49719 passed, 1156 skipped

---

Merge workflow run.