Full sync need to broadcast block hashes to peers after a sync cycle completed

[jangko]

This is mentioned in the spec albeit a bit vague and not forced. Also needed in hive test.

preparation:

• maintain a knownByPeer data structure. Can be as simple as Table[Peer, Table[BlockHash, Time]].
• periodically, cleanup this table after a period of time have passed(e.g. every 20 minutes)
• although this table is primarily used by newBlockHashes and newBlock handler of eth wire protocol, we need to prepare for it.

on peer connected(not mentioned by the spec, but implicitly by hive test, maybe part of robustness test):

• if peer bestDifficulty is lower than our total difficulty:
• using peer bestBlockHash to calculate distance with our best block hash.
• if the distance is one, send newBlock
• if the distance is more than one but less than e.g. maxHeadersFetch, send newBlockHashes.
• if the distance is too far, ignore the peer

after a sync cycle completed:

• get peers from peer pool(or local peer pool maintained by the synchronizer), except the current peer.
• build a hash list filtered by knownByPeer for each peers.
• check if the block is POS block or not. if it it a POS block, do not broadcast it.
• if the hash list is not empty, send it to the peer using newBlockHashes.

on peer disconnected:

• remove peer from knownByPeer table

[jangko]

Obsolete after full sync removed from nimbus-eth1