Cleaned up some trailing white space, converted hard tabs to soft, and added a
few missing _r8 modifiers

Stand-alone build found some unused variables and imports that I removed

Also removed trailing whitespace I inadvertantly added to marbl_diagnostics_mod
in the last commit

Some settings in marbl_settings_mod.F90 only need to be defined if
lvariable_PtoC (or NtoC) is true. Also, found a couple block in
marbl_init_tracer_metadata_mod.F90 that wasn't verifying N_ind > 0 before
modifying N tracers (N_ind = 0 if lvariable_NtoC = .false.)

For cesm2.0 and cesm2.1 (3p1z or 4p1z), add lvariable_NtoC but set the default
to .false.; also renamed gQfe_0 -> gQfe_max and added the non-N settings from
the settings_latest files.

Don't know what to use for Qn_fixed, waiting to hear back from UCI and NOAA
folks... so this will still fail the init.py test (but with fewer MARBL ERROR
lines)

Also generate the 4p2z settings file from JSON as part of testing

After discussion with Keith L, it looks like Qn_fixed and Qn_zoo were added to
the code base so the fixed N:C configuration looked more like the fixed P:C
configuration... but it isn't necessary since we want to use the same fixed N:C
throughout the code (as opposed to fixed P:C, which really does vary from among
the different functional types even when fixed spatially / temporally)

Modify marbl_io_mod to initialize N tracers to C tracers * 16/112 and set
feredsedflux and feventflux = 0.
Modify marbl_diagnostics_mod.F90 to include dust%remin*dust_to_Fe in integrand,
not righthand-side of iron conservation check
Modify marbl_interior_tendency_mod.F90 to sflux_in * desportion in P_iron%remin
before computing sflux_out.

Also cleaned up a few small things in interior_tendency_mod (use unit_system
instead of assuming cgs; initialize Lig_prod and declare it as inout in some
routines)

no coccoC in this initial condition file, but coccoC = 0.07*spC