If a pruning node ends up in a state where it has very few blocks on disk, then a node could fail to start up in VerifyDB. This pull changes the behavior for pruning nodes, so that we will just not bother trying to check blocks that have been pruned.
I don't expect this edge case to be triggered much in practice currently; this is a preparatory commit for segwit (to deal with the case of pruning nodes that upgrade after segwit activation). @sipa