When timeseries_insert.js is run in the parallel test suite, it re-uses the collection name from previous tests in the same thread. To avoid issues with existing unique indexes, this test should drop the collection before repopulating it with data.