Pre-test with Caution: Event-Study Estimates after Testing for Parallel Trends