sending backoff to tap producer when disk write queue > 1M is not optimal for large cluster with> 100M items and lot of RAM

right now we stop doing more rebalancing/replicating if the destination node's disk write queue has more than 1 million items

this number doesnt fit well if user has hundreds of million of items in the cluster .
for instance if you have a cluster with 120 million items it means each vbucket has 600k items and even moving one bucket at a time will cause the disk write queue to be more than 1M immediately considering ec2 disk performance

also some users they have deployments with enough RAM that they can keep more items in memory than others.

so disk write queue limit should be decided based on number of items in the cluster
amount of RAM and other factors.