salinity units are a mess when move to TEOS10

[StephenGriffies]

Salinity units need to be reconsidered when using TEOS10. There are various unsatisfying issues, such as the following:

--When MOM6 uses practical salinity, the diagnostics in fact state that the salinity units are ppt.

--PPT units are not CMOR recognized: need to change to 1e-3.

--PSU are not recognized units in CMOR.

--CMIP6 requests practical salinity to be archived, not absolute salinity. So when running TEOS10, with absolute salinity the prognostic field, we need to archive practical salinity, not the model's prognostic salinity field.

[Hallberg-NOAA]

This is still an outstanding issue that should be addressed even before a move to TEOS10.