Indexer rebalance should not choose a leader that is being removed.

Right now index rebalance will randomly choose the leader of the index rebalance. Which can be a node that is being removed from the cluster. It is the leader logs that holds information about the rebalance. This means that we can lose important debug information about the index rebalance.

At the very least a node that is remaining in the cluster should be choose.

Later we should look at replicating the logs across nodes (A new MB should be opened for this).

I have marked this as a bug as this breaks our FTDC model.