fix(playwright): reset cleanup state at start of _afterSuite to prevent worker crashes

[gololdf1sh]

Problem

When running tests with run-workers, workers can crash silently due to stale cleanup state leaking from test-level hooks into _afterSuite().

Root cause

1. A test fails → the screenshot plugin calls saveScreenshot()
2. If the screenshot also fails (timeout, closed page, protocol error), saveScreenshot() sets this.hasCleanupError = true and pushes to this.testFailures[] (Playwright.js L2991-2992)
3. The error is caught and the method returns gracefully — the test runner continues
4. hasCleanupError / testFailures are reset per-test in _beforeTest() (L567-568), but not at the start of _afterSuite()
5. When the last test in a suite has a screenshot failure, the stale state carries into _afterSuite()
6. At the end of _afterSuite(), the guard clause if (this.hasCleanupError && this.testFailures.length > 0) evaluates to true and throws (L841-844) — even if the suite's own cleanup completed successfully
7. In a run-workers context, this throw kills the worker thread. All remaining test files assigned to that worker are silently dropped — they never execute and never appear in the results

Fix

Reset hasCleanupError and testFailures at the top of _afterSuite(), before any suite-level cleanup runs. This ensures the guard clause at L841 only evaluates errors from _afterSuite's own cleanup operations, not leftover test-level errors.

  async _afterSuite() {
+   // Reset leftover test-level cleanup state (e.g. screenshot failures)
+   // so only errors from this suite teardown are evaluated below.
+   this.hasCleanupError = false
+   this.testFailures = []
+
    // Stop browser after suite completes