uwshcu.F90 has a regression bug

[atrayano]

The limiter uses a scheme involving cnvtr. But the pointer to this variable is passed in from an export of Moist (CNPCPRATE).

The regression (start/stop and 0 IAU increments) fails, unless tests are using either LM=72, or BACM_1M, or UW_OPTIONS: NONE

Inside GEOS_UW_Interface.F90 there is logic that if (LM==72) the default for JASON_UW is set to true (false otherwise). And with JASON_UW on, one of the parameters we set there effectively chooses a different limiter. Of course, the regression passes when we choose BACM_1M since we are not using UW shallow water code,