make_coupler_mosaic error in area_atm field.

[ngs333]

make_coupler_mosaic_parallel (and possibly make_coupler_mosaic) is not correctly calculating or storing the area_atm field. The other fields (including area_lnd) seem to be correct. This is for at least for when the input is that of Test04.
Here are some additional observations and questions:
a) running with different number of mpi ranks yields different results for that field.
b) Test04 is not always detecting such failures when it is run.
c) Values for the field are often non-sensical, e.g. from e+220 to e-318, numerous are exactly 0. Some sections appearing random.
d) It may be interesting to know when this issue first appeared in the baseline.
f) It may be useful to know the consequences of calculating this particular field (area_atm) incorrectly.
g) the app does print some warnings, and these should be noted.

To Reproduce
Essentially run the tail end of Test04 (the calls to make_coupler_mosaic_parallel) with baseline code on
Analysis, and preferably with code compiled with the canonical build method.
Then compare with nccmp the land_mask_tileX.nc files in the /parallel directory and the one above it.
Note the discrepancies. Also run the serial version and compare the resultant files to those other .nc files produced by the serial program.

Expected behavior
There should be no difference in the land_mask_tileX.nc files.

System Environment
This can be observed on Analysis, building with the current canonical method (using the site-config, env.sh file, which currently loading for gcc 11)