Merge branch 'rljacob/cime/update2-mct-2.10.beta' (PR #1452)

Update MCT with performance improvements from Pat and Az. These are needed for high-resolution simulations.

Had to remove and re-add subtree because directory location moved.

Threading and vectorization updates to AttrVect and MatAttrVectMul
Disable RSEND by default
Decrease serial complexity of initp_ algorithm to speed it up.
Speed up peLocs at the cost of some increased memory.
Import swapm implementation of MPI_AlltoallV from PIO and use in Rearranger
Reduce complexity of active_pes_
Also add some mpi-serial features

[BFB]

* rljacob/cime/update2-mct-2.10.beta:
  Allocate send/recv buffs even if nprocs eq 0
  Allocate send/recv buffs always when useswapm
  Squashed 'cime/src/externals/mct/' content from commit 72c0d28
  Removing cime/src/externals/mct for subtree update

src/externals/mct/mct/Makefile

@@ -31,6 +31,7 @@ SRCS_F90	= m_MCTWorld.F90		\
 		  m_SparseMatrixPlus.F90	\
 		  m_Router.F90			\
 		  m_Rearranger.F90        	\
+		  m_SPMDutils.F90               \
 		  m_Transfer.F90
 
 OBJS_ALL	= $(SRCS_F90:.F90=.o)
@@ -82,6 +83,7 @@ m_GlobalMap.o:
 m_GlobalSegMap.o:
 m_GlobalSegMapComms.o: m_GlobalSegMap.o
 m_Navigator.o:
+m_SPMDutils.o:
 m_AttrVectComms.o: m_AttrVect.o m_GlobalMap.o
 m_AttrVectReduce.o: m_AttrVect.o
 m_AccumulatorComms.o: m_AttrVect.o m_GlobalMap.o m_AttrVectComms.o
@@ -91,7 +93,7 @@ m_GeneralGridComms.o: m_AttrVect.o m_GeneralGrid.o m_AttrVectComms.o m_GlobalMap
 m_MatAttrVectMul.o: m_AttrVect.o m_SparseMatrix.o m_GlobalMap.o m_GlobalSegMap.o m_SparseMatrixPlus.o m_Rearranger.o
 m_Merge.o: m_AttrVect.o m_GeneralGrid.o
 m_Router.o: m_GlobalToLocal.o m_MCTWorld.o m_GlobalSegMap.o m_ExchangeMaps.o
-m_Rearranger.o: m_Router.o m_MCTWorld.o m_GlobalSegMap.o m_AttrVect.o
+m_Rearranger.o: m_Router.o m_MCTWorld.o m_GlobalSegMap.o m_AttrVect.o m_SPMDutils.o
 m_GlobalToLocal.o: m_GlobalSegMap.o
 m_ExchangeMaps.o: m_GlobalMap.o m_GlobalSegMap.o m_MCTWorld.o m_ConvertMaps.o
 m_ConvertMaps.o: m_GlobalMap.o m_GlobalSegMap.o m_MCTWorld.o