Mongorestore does not retry writes on transient errors such as NotWritablePrimary, which can occur during an election. We should add retry logic, similar to mongosync, so that users don't need to manually rerun mongorestore on these errors. 

According to the documentation, the driver makes 1 retry attempt for network / failover errors. This is based on serverSelectionTimeoutMS, which defaults to 30,000 ms. However, we still saw mongorestore fail in HELP-69430 during an election.