When drivers talk to mongod, it caches the highest term it has seen for that cluster. If a node with a lower term tries to claim that its primary, the driver won’t accept it. This problem still exists for disagg, so the magic restore path needs to use the log service to insert a no-op at the top of the restored log with an arbitrarily high term greater than the last term we've seen.

Implementation was done in another ticket, this is just for testing it.