WL#2636: NbdCluster adaptions for merge-join

With implementation of 
WL#2591 Merge join behavior when two ordered indexes are available
some adaptions might be necessary for ndbcluster handler and ndb
ordered index scan (restarting or continuing open scans with new range).
Also any new query executuion statistics needed from ndb for optimizer
to decide for a merge-join belongs here.
This WL is opened for a HL discussion of what will need to be implemented
(if anything).