check_forcing() is flooding stdout in CMOM compset

[mnlevy1981]

#419 introduces a new function to make sure the column fractions provided by the GCM for PAR sum to 1 (within 1e-6); when turned on in a C compset testing with the MARBL driver in MOM, we are getting a ton of warnings about 1e-3 off, which is crippling performance.

The PR currently has an lcheck_forcing flag that is false by default, but once we figure out why the C compset fractions aren't adding up we should remove that flag (so we always call the function) and possibly also change the warning to an error.