ClusteredListenerReplTest.testPrimaryOwnerGoesDownAfterBackupRaisesEvent random failures

ClusteredListenerReplTest uses a BlockingInterceptor to ensure a write command is blocked on the backup owners while the primary owner is killed.

The BlockingInterceptor is removed after the primary owner was killed, but only on one of the backups. This is enough if the node 0 becomes the primary owner of the key, because the command is retried before the previous invocation is unblocked, and BlockingInterceptor only blocks one command at a time. But if node 2 becomes the primary owner, the retry happens after the initial command was unblocked, and it blocks in BlockingInterceptor.

The fix would be to remove the interceptor from both nodes.

I also believe the "block one command at a time" logic should be removed, because it makes tests too dependent on timing.