Don't compute log10(chl) for small chl

[mnlevy1981]

if chl <= chl_min, we already have log10(chl_min) stored as a parameter. Avoiding the computation of log10(chl) in those cases prevents the possibility of a floating point exception (if chl <= 0), and this approach replaces a max() statement with an if block so it shouldn't affect performance.

This requires adding chl_min to the optics_type (which already has log10chl_min and log10chl_max).

[mnlevy1981]

Note that this fix will allow us to run CESM with DEBUG=TRUE and CHL_FROM_FILE=False (if MARBL is available to provide CHL). Previously, debug tests were failing because MARBL was returning chl=0 at some points and MOM6 was dutifully trying to compute log10(chl) anyway. The MOM_interface tag that includes this MOM6 update should add an SMS_D test for C1850MARBL_JRA to aux_mom_MARBL (and maybe prebeta?) to make sure we can continue to run in debug mode if needed.

[mnlevy1981]

testing: Verified this is bit-for-bit in a one month test run (with MARBL providing chlorophyll)

[klindsay28]

@mnlevy1981 , do you know why this is needed in lookup_ohlmann_swpen, but not lookup_ohlmann_opacity, which has the same log10chl computation?
Perhaps @fobryan3 has an idea about this.