The integration tests were flaky because they relied on a shared booking
created in beforeAll, which could cause race conditions when tests run
in parallel or when cleanup from previous runs interfered.
Changes:
- Use seed user (member0-acme@example.com) when available, fallback to
creating a test user for local development
- Create a fresh booking in beforeEach instead of sharing one across tests
- Track created bookings and clean them up properly in afterEach
- Fix import organization to satisfy linter
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
