More thoroughly clean up test objects after each iteration of benchmark tests

In https://github.com/Azure/azure-functions-sql-extension/pull/439 I added some more cleanup such as deleting the leases table. It would be better to look into starting with a clean slate for each iteration and making a new database and/or table so that we're sure there's nothing that will impact successive runs.

Start by reviewing PR #439 and the benchmark test setup to understand the cleanup already added, including deletion of the leases table. Define the iteration lifecycle so each run creates a fresh database and/or table, with cleanup verified between successive runs.

Written by the indexing model from the issue text.