Make E2E authentication setup resilient

[isaacroldan]

Why

Remote E2E authentication failures do not identify the failed stage, and transient device-login or page-load failures require a complete rerun.

Stacked on #8325.

What

• Validate required remote E2E configuration when global authentication starts.
• Classify authentication failures as configuration, PTY startup, device-code generation, browser login, or session prewarm.
• Retry only transient device-code, login page-load, CLI confirmation, and prewarm page-load failures once.
• Remove the per-worker fallback login so all remote workers use the diagnosed authentication setup.
• Emit safe stage and reason diagnostics without raw CLI output, login URLs, credentials, cookies, or load-test header values.
• Wait for delayed client-side redirects before validating authenticated prewarm destinations.
• Add local Playwright coverage for validation, retry selection, retry boundaries, HTTP failures, destination waits, and diagnostic redaction.

Testing

• pnpm --filter @shopify/e2e lint
• pnpm --filter @shopify/e2e type-check
• pnpm exec playwright test tests/auth-diagnostics.spec.ts --project local (9 passed)
• Workflow YAML parsing and git diff --check passed
• Both remote shard test lists include the authentication setup