Prevent Observer NameNode from becoming StandBy NameNode

When the cluster performs a failover from NN1 to NN2, NN2 is asking all the other NNs to cede active state and transit to StandBy including the Observer NameNodes.

Seems we should block Observer from becoming StandBy and participating in Failover. Of course, since we can transit StandBy NameNode to Observer, we can separately support promote Observer NameNode to StandBy NameNode.