Benchmarking: Ensure consistent housekeeping across benchmarking runs

From the experiments we've been running, we've observed sometimes conflicting results from testing the same configuration.

For example, #6930 (comment 2840457135) contains different results than !8207

One variable that is not being controlled for is housekeeping. Let's ensure that housekeeping is run in a consistent manner across all benchmarking runs so don't see variance due to varying repository state.