Plasma: Allow garbage collection when a snapshot is being iterated upon

In Eventing, we create snapshots during rebalance to transfer doc timers mapped to specific vbucket(s) to another eventing node. For that on source eventing node, we create snapshot on plasma instance and filter out contents related to vbucket supposed to be transferred. In this case, memory usage of eventing process on source node grows up quite a bit during rebalance - so a facility to control the memory footprint related to plasma would help.