Don't lose ETCD watch updates during cluster inaccessibility

[fit51]

Current implementation - loads all key-value pairs during initialization, subscribes on changes and this is it.
For better fault tolerance I propose to add storage update activity.
This activity would periodically load all key-value data since last revision, update storage TrieMap send changes events if it finds stale data.

Possible scenario

• Node loses network connection with ETCD cluser. WatchService starts to reconnect.
• User update keys 1 and 2 on ETCD.
• Network connection is restored. Node succeeds to resubscribe on Watch. Now it would receive new changes, but it has lost 1 and 2 keys update.

Problems to solve in this proposal

• Possible update duplicates. One could come from watch stream, another - from storage update activity.