Management of long forks

[albertog78]

In current implementation, if node A is synced up to a certain block and node B, being on a fork started NetworkSettings.desiredInvObjects + 1 blocks before A tip, node B even if "understand" the exact branching point of node A, will send just NetworkSettings.desiredInvObjects ids of its active chain. In such situation A, even after having received these new blocks, will publish the new SyncInfo without any information about these new blocks (node A is still on the old branch) and for this reason B will again send to A the same NetworkSettings.desiredInvObjects blocks already sent at the previous iteration. This will be a never ending loop preventing the two nodes to converge to the same tip.
Using the locator (and not just the simple contiguous list of ids) will even cause the problem more easily.

[i-Alex]

Also review MainchainSynchronizer class