This creates a unique bucket name for each test, so that the tests in parallel don't collide, and buckets left over from interrupted tests don't cause future failures. Also make sure that buckets are removed, regardless of content.