Currently, we seem to use upgradeCluster to double as downgrade by passing in a downgraded version (example). However, our documented downgrade procedures indicate that the order in which we restart the binaries should be the reverse of the upgrade procedure:
For upgrade, we restart in order of config servers -> shards -> mongos
For downgrade, we restart in order of mongos -> shards -> config servers.

For this ticket, we should create a separate downgradeCluster helper and make sure each reference to upgradeCluster(downgradeVersion) is updated appropriately.