Make retryable_findAndModify_commit_and_abort_prepared_txns_after_failover_and_restart.js not test restart with a 1-node shard

Due to the issue in described SERVER-62951, a 1-node replica set may end up performing replication recovery in the node is state "primary" rather than "secondary", and this can lead to a deadlock when there is a prepared transaction to reconstruct.