In nuopc cap, handle decompositions that don't include all points

[billsacks]

@whlipscomb introduced an improved decomposition strategy that excludes ocean regions from the decomposition. We put a change in the MCT cap to handle this (3fdb962). A conceptually similar change is probably needed in the nuopc cap, possibly following what @mvertens did for the CTSM decomposition (which involves adding ocean cells at the end of each processor's owned points).