Add a tests that sets period for refresh method to 100 ms and runs core suite on 3 node replica set.
This suite will likely uncover bugs in logical sessions cache caused by race conditions.

Approach:
1. Change the parameter logicalSessionRefreshMinutes to logicalSessionRefreshMillis, so that we can drill down to the millisecond granularity.
2. Update the shell parameters to reflect logicalSessionRefreshMillis. Consider if we want to backport this change to 4.0/3.6.
3. Change the parameter kLogicalSessionDefaultRefresh to 300000ms instead of 5 minutes.
4. Create a passthrough suite that uses a 3-node replica set (sharded cluster?), a logicalSessionRefreshMillis value of 100ms, and runs the core suite.