Numerical derivatives for Monin-Obukhov exchange coefficients over land in offline mode

[gmao-rreichle]

Supersedes #918. Adds optional numerical derivatives of Monin-Obukhov exchange coefficients over land in offline mode. Numerical derivatives implemented for Louis and Helfand. Also cleans up and documents surface turbulence code in Catchment GridComp. Numerical derivatives not yet implemented for CatchCN.

New resource parameter MOSFC_EXTRA_DERIVS_LAND replaces incl_Louis_extra_derivs:

     ! MOSFC_EXTRA_DERIVS_LAND: Resource parameter for *offline* (LDAS) mode.
     !
     !  Over *land*, use derivatives of exchange coeffs w.r.t. temp. & humidity.
     !    
     !    0 : None,               default for Helfand.  
     !    1 : Analytical derivs,  default for Louis,    *not* available for Helfand.
     !    2 : Numerical  derivs. 
     !    3 : Numerical  derivs,  via virtual temp.,    *not* available for Helfand,  same as 2 but faster than 2.
     !
     !  Runtimes: Helfand takes ~10 times longer than Louis.  In offline mode, Helfand consumes
     !            about as much CPU as Catchment.  Option 2 triples the runtime of the MOSFC scheme.
     !            Option 3 doubles the runtime of the Louis scheme.  

Successfully 0-diff tested for GEOSldas Intel standard tests at 3b6cced

[biljanaorescanin]

@gmao-rreichle here is testing summary:

1. All GEOSldas regression tests passed.
2. Regression tests for AGCM all failed for change of long name.
   Fail is same for all 1 day AMIP, REPLAY, And no inc REPLAY.
   Here is example from AMIP:
   Comparing NC4 catch_internal_checkpoint using nccmp...
Failure!
Checking for data differences
Checking for metadata differences
DIFFER : VARIABLE : CQ : ATTRIBUTE : long_name : VALUES : "surface_moisture_exchange_coffiecient" <> "surface_moisture_exchange_coefficient"
3. Running regression script was a pass.

[biljanaorescanin]

I've redone tests since it has been a while:

1. GCM AMIP, REPLAY and INC REPLAY all have just this difference in change of the long name:

Comparing NC4 catch_internal_checkpoint using nccmp...
Failure!
Checking for data differences
Checking for metadata differences
DIFFER : VARIABLE : CQ : ATTRIBUTE : long_name : VALUES : "surface_moisture_exchange_coffiecient" <> "surface_moisture_exchange_coefficient"

2. Regression tests for GEOSldas: GLOBACS/assim test fails to run unless the rc variable MOSFC_EXTRA_DERIVS_OFFL_LAND: 0 is added to the test's "exeinp" config file (globalcs.assim.exec). With this change, the test passes.
   All other tests were zero diff.

[biljanaorescanin]

@sdrabenh this is ready for you
@mathomp4 once merged we will have to add rc variable MOSFC_EXTRA_DERIVS_OFFL_LAND: 0 to test's "exeinp" config file (globalcs.assim.exec).