* save funding_locked_received: if a node already sent us
funding_locked, save it to avoid superfluous messages
* use Queues instead of Futures: this ensure that we don't error if we
receive two messages of the same type, and in avoids having to delete
futures in finally blocks. A queue monitor could be added to detect
queue elements that are not popped.
* request initial routing sync: since we don't store the graph yet, it
is better to request the graph from the Peer so that we can route
* channel_state cleanup: now each channel should have a state, which is
initialized to OPENING and only marked OPEN once we have verified that
the funding_tx has been mined
