Sharding improvements

[jait91]

• refactor perf test
• remove unnecessary round timer from round managers
• add finalized flag to blocks and add recovery mechanism on startup (retry storing commitments to DB and finalize the block)
• fatal exit on processing errors - exit immediately on round processing failures, recovery handles on restart

[lploom]

LGTM

[b3y0urs3lf]

Tested locally using GO aggregator commit b271ce2.

1. It is possible to start aggregator in sharding mode (2 shards), and successfully submit commitments to it (commit + get inclusion proof).
2. It is possible to start aggregator in sharding HA mode (2 shards), and successfully submit commitments to it (commit + get inclusion proof).

3. Tested that it is possible to execute perf test on top of sharding in HA mode

4. Also checked that after droping mongoDB container related to specific sharding node, then this node would crash and when mongoDb container would be UI it would be able to establish connection to it (tested in sharding mode without HA). 5. ALso I was able to sucessfuly add sharding config to subscription sevice And I was able to send successfully commitments through subscription service.

Merging and moving task to done.