Don't send expedited blocks until PoW verified

[kyuupichan]

At present we send an expedited block before checking the PoW or merkle root. This is a bad idea, but worse it means our peer will ban us if we send something bad, and without the above checks it's easy for an attacker to get a BU node to send an invalid expedited block.

This patch remedies that by not sending one until both checks are done.

[kyuupichan]

I've just realised there is no point in the inv in thinblock.cpp. Let me resubmit.

[kyuupichan]

Re-pushed without pointless Inv.

[ptschip]

@kyuupichan I understand the intent here but I'm not sure we can do this...checking the merkle root for expedited processing will take too much time if re-requests are needed and will also add a fair amount of time just in checking for collisions and verifying we have all the tx hashes , and will thus defeat the purpose of expedited forwarding. @gandrewstone what do you think? you know more about expedited performance...

As for bans on the expedited network , I think expedited nodes should be all whitelisting each other anyway, but we still need a bigger discussion (as you've mentioned) about creating a proper policy for when , when not to ban nodes.

[kyuupichan]

@ptschip I want expedited to be the ultimate default method of operation to extract the maximum gain xthin has to offer. We cannot be selecting trusted peers and whitelisting them. It has to be good for any peer to be an expedited peer, otherwise something is wrong.

We can't be sending invalid stuff around the network. bitcoind was always very careful about that, and for good reason. Validation is an unfortunate necessity.

[kyuupichan]

Updated to only check PoW which is nearly instant. Merkle root checking is deferred and banning removed.

[kyuupichan]

Rebased to avoid the conflict now #482 is in.