Full backfill of replicas after any restart

Just two nodes in cluster_run. I've tried with 1k vbuckets and with just 32 (to have more items). 1600 k of bucket quota.

• Loaded 1.2M of 1k objects (all zero). It'll be slightly DGM

• Waited until replica count and item count is ok and until everything is persisted.

• Waited 10-20 second more.

• killall -SIGTERM memcached.org

Observe that replicas are rebuilt:

rebalance:info] [2012-03-01 15:31:52] [n_1@10.17.2.163:<0.1597.0>:ebucketmigrator_srv:process_upstream:354] Initial stream for vbucket 27

and see that replica items count grows from 0 to 1.2 M and write queue is non-zero for some time.