[Design] Add optimize strategy for message dispatching in network

[fbzhong]

In current implementation of go-nebulas, a network message, for example, a newtx message will broadcast to whole network, each network will also relay to peer nodes as well. That makes broadcast storm effect, which still has some impaction even if we implement receved_message strategy.

When a duplicated message are spread in network, they bring lots of computation pressure to blockchain core, especially newtx message during stress test.

To optimize that, I suppose to introduce a duplication message check in dispatcher.go module.

In function PutMessage(), check whether this message are dispatched. If so, ignore; otherwise escalate to upper level.

[fbzhong]

Update:

add metrics for duplicated messages in commit 4d843a0

[fbzhong]

So far, this strategy works, and the following commits are related to this change:
5adbb32 net: dispatcher.go. support enable & disable duplicate msg filter for…
d6cc4e4 core: block_pool.go. remove filter from downloaded block message.

The reason why we introduce this strategy is the broadcast storm, we should find new strategy to prevent that.

[fbzhong]

#56 seems a bug related to this strategy.