Enhance view reliability by allowing for replica reads

At the moment, when any node is unavailable in the cluster, any view request will be forced to return incomplete data until it is failed over.

Given that we have replica-read capabilities for individual documents, and given that the replica views are already in place...can we add the ability to query that replica view data? Initially it could certainly be a manual option as is the replica-read for documents but it would also be reasonable to make it an automatic step if the view response times out to one or more nodes.