stablerange: use sort cache to build 'mergepoint' information Using the walk cache gives use a better complexity we need to scale. There are further improvement we could make, but this is a good first step.

stablesort: make the iteration from head available to all This will be useful for the stablerange code to work properly.