obs_def_land_mod and soil moisture

Use case

The observation converters for soil moisture observations are not consistent. We may need to specify more DART QUANTITYs to cover the situation. Many NASA products use 'volumetric soil moisture percent', CLM uses both
'mm3/mm3' and 'kg/m2', NOAH uses 'm3/m3', so it is really difficult to write a forward operator to return
what is expected, and similarly the observation providers use an equally disparate set of units. Conversion between
many of these requires knowledge of soil properties.

Depending on what variables people use to include in the DART state, the units of the returned quantity are different.
If the CLM history file is used to provide H2OSOI as QTY_SOIL_MOISTURE, the units are mm3/mm3:

	float H2OSOI(time, levsoi, lat, lon) ;
		H2OSOI:long_name = "volumetric soil water (natural vegetated and crop landunits only)" ;
		H2OSOI:units = "mm3/mm3" ;```

If the CLM restart file is used to provide the prognostic variables H2OSOI_LIQ, H2OSOI_ICE, the units are kg/m2:

	double H2OSOI_LIQ(column, levtot) ;
		H2OSOI_LIQ:long_name = "liquid water" ;
		H2OSOI_LIQ:units = "kg/m2" ;

	double H2OSOI_ICE(column, levtot) ;
		H2OSOI_ICE:long_name = "ice lens" ;
		H2OSOI_ICE:units = "kg/m2" ; 

So - if people specify H2OSOI from the history file as the source of QTY_SOIL_MOISTURE, they get soil moistures appropriate
for the obs_def_COSMOS_mod.f90 and maybe some other observations. The CLM model_mod currently tries to return H2OSOI if asked for a soil moisture and if it does not exist in the state, the soil moisture returned is the sum of H2OSOI_LIQ and H2OSOI_ICE .. which has units of kg/m2 and is NOT appropriate for the obs_def_COSMOS_mod, but is useful for
other soil moisture observations.

Is your feature request related to a problem?

We need to assimilate soil moisture observations - mostly from SMOS, SMAP - satellite observations.

Describe the your prefered solution

I don't have a preferred solution. More DART QUANTITYs? Force the obs_converters and model_interpolate routines to
return identical units?

Describe any alternatives you have considered

The CLM model_mod supports different units by including different variables in the DART vector. This is incredibly cryptic
and prone to error.