Currently, when we try to restart an empty standalone, it will fail trying to recover, and we fail it with an fatal assertion.

Additionally, when we try to restart a replica set node with the recoverFromOplogAsStandalone parameter set, it also fails. We fail it with an invariant here.

We should consider changing the fassert and invariant to fassertNoTrace to avoid core dumping and stack tracing, since the issue is with the startup parameter/the data the server has been fed, and not an issue with the server logic itself.