SERVER-102818 added an extra suite to add and remove shards while running a fsm workload. As part of this, the movePrimary state had to be changed to check and retry a number of times if the destination shard was removed. However, when giving up, there is no return true; statement which would effectively stop the assert.soon execution.

The purpose of this ticket is to fix this behavior, by either adding the statement, or, re-evaluating the usage of assert.soon, considering that as this is an fsm, eventually the movePrimary state would be re-executed, and simply adding the ShardNotFound error as a permitted failure for movePrimary. If we want to be more strict, we could add the error only if we are running in an add/remove shard suite.