Similar to network_error_and_txn_override.js, run_reconfig_background.js should handle ReplicaSetMonitorErrors where we cannot find the primary because it has stepped down or been killed.

We should also specify nodb for the DoReconfigBackground hook so that run_reconfig_background.js can explicitly try to connect to mongo. If we don't do this, it is possible for the test infrastructure to try to connect to mongo before a test has started.