Currently teardown of both replicaset and sharded cluster fixture are done serially.
This can take several seconds and could be speed up by just parallelizing the shutdown of the different processes in the cluster.

Speedup:

• teardown of replicaset fixtures: 3x (3 secs -> 1 secs)
• teardown of sharded cluster fixture: 4.3x (13 secs -> 3 secs)