Fix two issues that can cause a tassert or crash if a stepdown happens during authoritative collection metadata refresh:

1. If a stepdown interrupts the refresh, the refresh may swallow any error code and keep retrying. Eventually it will hit the retry limit and tassert.
   • We should stop retrying if the opCtx has been interrupted.
2. On stepdown the replication coordinator interrupts all optime waiters and then invariants that none remain. However we are missing those registered by ReplicationCoordinator::registerWaiterForMajorityReadOpTime (which are used when recovering the authoritative metadata from disk).
   • We should also interrupt those optime waiters.