Tree-service sync after splitbrain failover

[anatoly-bogatyrev]

Incorrect tree-service sync after splitbrain failover.

Synchronization occurs only from one node.
In this case, the node could synchronize data with a node from the same part of the cluster.
The order for the container is fixed.

Need to fix this:

• get not fixed node to sync (can be selected sequentially or in random)
• sync not only after new epoch event

[fyrchik]

Let's add randomization for now and discuss other ways to improve the syncronization.

[alexchetaev]

Need to create issue for discussion