Make configurable and increase default value of timeout used to check memcached prior to TAP -> DCP change

When upgrading from 2.x -> 3.x, after the last node becomes 3.x we commence the TAP -> DCP upgrade (which is essentially a rebalance.) Before starting the upgrade we quickly check the health of the memcached nodes in the cluster to ensure their health. This is done with a call to get vbucket states with a non-configurable timeout of 1s. One of our customers is running into the case that this call times out and prevents them completing the TAP -> DCP upgrade. Fixing this will at least allow them to move forward past this step.