description of repl cache access contains outdated information and needs update

The chapter "request-balancing" is not correct as any replicated cache will work similar to distributed (exception is that numOwner==numNodes).
So it is aware of ConsistentHash and the client will route the request always to the primary owner based on CH.
There is no round-robin for normal access.

Only in case of cluster-view changes or node failures the retry might behave different.