Bug without consequences in call to gwpds.f from cires_ugwp.F90

[climbfuji]

The crooked way of calling gwdps_run from cires_ugwp_run, i.e. from a different CCPP scheme, contains a bug that fortunately doesn't have any consequences.

In cires_ugwp_run, we are passing the entire qgrs array with dimension (im,km,ntrac) to gwpds_run. gwpds_run, however, only expects a 2D array for specific humidity of size (im,km). This works by chance, because specific humidity is stored in the first slice of qgrs, i.e. in qgrs(:,:,1).

Thanks to @mdtoy for finding and reporting the bug.

[climbfuji]

Fixed in #483.