Pre-test with Caution: Event-study Estimates After Testing for Parallel Trends