Is your feature request related to a problem? Please describe.

currently, loss of any of the split-chain members (shards) of user data makes it inaccessible. For REP X policies, this happens when all X replicas are unavailable

on the other side, user data from EC containers (#3420) is encoded with redundancy so that it can be restored event if up to R data shards is corrupted or even completly unavailable

but to make this happen, it is necessary to support corresponding recovery in GET server

Describe the solution you'd like

extend object GET assembly algo. Once it detects some data shard is unavailable, it:

1. resolves [1, ..., K+R] RS chain using already got shards or SEARCH API
2. pull as much shards as possible, but not less than K
3. restores missing data shards and return them