Next will be to create marbl_interior_mod and pull out the interior routines.

Note that this commit does not build - other modules are still trying to use
marbl_mod, which no longer exists.

marbl_glo_avg_mod: has marbl_glo_avg_count_vars and
                   marbl_glo_avg_init_rmean_vals (both called during init()) as
                   well as module variables used by next two modules
marbl_surface_mod: has set_surface_forcing
marbl_interior_mod: has set_interior_forcing and
                    marbl_set_global_scalars_interior as well as lots of small
                    private functions called by those two

Note: need to come up with better names for functions in marbl_surface_mod and
marbl_interface_mod. One idea is to rename the modules again:

marbl_surface_mod -> marbl_surface_fluxes_mod [with
                     marbl_surface_fluxes_compute]
marbl_interior_mod -> marbl_interior_tendencies_mod [with
                      marbl_interior_tendencies_compute]; unsure of
                      marbl_set_global_scalars_interior, though. Maybe
                      marbl_interior_set_global_scalars? (maybe rename
                      functions but not the modules?)

Also add a use statement to set_interior_forcing(); missing
marbl_diagnostics use statement leads to linking error though libmarbl.a
builds fine.

interface: set_surface_forcing() -> compute_fluxes()
surface: marbl_set_surface_forcing() -> marbl_surface_compute_fluxes()
ciso: marbl_ciso_set_surface_forcing() -> marbl_ciso_compute_fluxes()

interface: set_interior_forcing() -> compute_fluxes()
interior: marbl_set_interior_forcing() -> marbl_interior_compute_tendencies()
ciso: marbl_ciso_set_interior_forcing() -> marbl_ciso_compute_tendencies()

Function is now marbl_nhx_surface_emis_compute(), conforming to the convention
that all public functions in marbl_XYZ_mod have a name marbl_XYZ_[something]()

It is now marbl_surface_flux_mod, and marbl_surface_compute_fluxes() is now
marbl_surface_flux_compute(). Similarly, marbl_interface%compute_fluxes() is
now marbl_interface%surface_flux_compute().

It is now marbl_interior_tendency_mod, and marbl_interior_compute_tendencies()
is now marbl_interior_tendency_compute(). Similarly,
marbl_interface%compute_tendencies() is now
marbl_interface%interior_tendency_compute().

Moved to marbl_ciso_surface_flux_mod, and next commit will split it to also
create marbl_ciso_init_mod and marbl_ciso_interior_tendency_mod

Now resides in marbl_ciso_init_mod; also created
marbl_ciso_interior_tendency_mod although it is empty as of this commit.

This completes the split of marbl_ciso_mod. I noticed that
marbl_interior_tendency_mod and marbl_ciso_interior_tendency_mod both had
similar update_particulate_terms_from_prior_level() routines that called
identical marbl_update_sinking_particle_from_prior_level() routines, so I moved
the latter to a new module (marbl_interior_share_mod) amd renamed it
marbl_interior_share_update_sinking_particle_from_level_above()

flux_as and flux_sa were both set but then not used in
marbl_ciso_compute_fluxes()

marbl_ciso_surface_flux_compute() and marbl_ciso_interior_tendency_compute()
both return immediately if ciso_on is false, so marbl_surface_flux_compute()
and marbl_interior_tendency_compute() can both call those subroutines without
checking the value of ciso_on first.

Also, I moved two routines that pack share data types to
marbl_interior_share_mod.F90 -- they get called regardless of ciso_on but
return immediately if ciso_on is false.

There are still a few places where we check the value of ciso_on that I would
prefer to avoid, but that will require a little more refactoring to make
possible.

Created marbl_ciso_diagnostics_mod.F90 to handle the ciso diagnostics as well
as marbl_diagnostics_share_mod.F90 to house subroutines / datatypes that are
used by both ciso and the base tracers.

nag caught three unnecessary use statements in
marbl_ciso_diagnostics_mod.F90

Always call marbl_ciso_init_tracer_metadata, return immediately if ciso_on is
.false.

This may not be the final name change, I think we need to figure out a better
name than "surface flux output" (it's confusing because surface_tracer_flux is
the primary output of surface_flux_compute() and also Chl isn't a flux but is
stored in this structure)

num_elements_surface_flux is a better name for the number of columns being
computed simultaneously inside surface_flux_compute().

We don't use _tracer in the marbl_surface_flux_mod name or any subroutines, so
I'm removing it from the variable as well. I also removed the "stf" shorthand,
just using surface_fluxes everywhere.

Also added a comment that this datatype is designed to pass data from
surface_flux_compute() to marbl_diagnostics_mod.F90

1. marbl_diagnostics_set_surface_forcing() is now
   marbl_diagnostics_surface_flux_compute()
2. store_diagnostics_ciso_surface_forcing() is now
   marbl_ciso_diagnostics_surface_flux_compute()

Since the check to see if ciso tracers are enabled is now in this routine, we
can not specify surface_input_forcings(surface_forcing_ind%d13c_id)%field_0d in
the argument list (because non-ciso runs will try try to pass
surface_input_forcings(0)%field_0d, which is out of range).

Lots of comments and a few log messages still used "surface forcing" rather
than "surface flux". As of this commit, I believe that all references to
surface_forcing refer to the forcing fields requested by MARBL for the
surface_flux_compute() routine.

Next commit will be to rename surface_input_forcings to just plain
surface_forcings as that is no longer an [incorrectly] overloaded term.

Now marbl_interior_tendency_share_type (matches marbl_surface_flux_share_type).
Also, local variables of this type are now marbl_interior_tendency_share

It is now interior_tendency_forcing_ind (same with indexing_type).

dtracers, column_dtracers, etc are all now interior_tendencies. The subroutine
compute_dtracer_local() is now compute_local_tendencies().

Variable is now interior_tendency_forcings

Now num_interior_tendency_forcing_fields

All occurences are now interior_tendency_forcing

Now tracers and tracers_at_surface, respectively.

All modules have a module-level "implicit none" so individual subroutines in
the modules do not need it.

Expect 'interior_tendency' instead of 'interior' and 'surface_flux' instead of
'surface'

Renamed to autotroph_zero_consistency_enforce and also moved the ciso check to
marbl_ciso_interior_tendency:marbl_ciso_autotroph_set_to_zero

* Renamed marbl_interior_share_mod to marbl_interior_tendency_share_mod
  (also updated public subroutines in the module)
* Changed ciso_on check in share modules from "if (.not. ciso_on)
  return" to "if (ciso_on) then" blocks [more extensible once we add
  other tracer modules]
* Created marbl_surface_flux_share_mod.F90 with an export_variables
  function to remove ciso_on from marbl_surface_flux_mod.

This removes ciso_on from marbl_interior_tendency_mod

marbl_interior_tendency_share_update_sinking_particle_from_level_above() is too
long for a function name (NAG warns about more than 63 characters and gfortran
throws an error (possibly because we elevate warnings to errors).

New name is marbl_interior_tendency_share_update_particle_flux_from_above()

num_elements_surface -> num_elements_surface_flux
num_elements_interior -> num_elements_interior_tendency
num_interior_tendency_forcings -> num_elements_interior_tendency

set_global_scalars (from marbl_interface_class) can call adjust_bury_coeff
directly. Now named marbl_interior_tendency_adjust_bury_coeff, the subroutine
returns immediately if ladjust_bury_coeff is .false.

Now marbl_glo_avg_var_cnts_compute()

Variables and datatypes both relied on surface_saved_state and
interior_saved_state rather than more descriptive names.

Now marbl_interior_tendency_diag_ind; also, included _tendency in type name

Instead of marbl_diagnostics_share_mod, these belong in
marbl_interface_private_types.

This should be the last of the code clean-up commits following review of marbl-ecosys#308

f2ff0b1 updated marbl_glo_avg_mod.F90 but I didn't change the names of
interface variables.

This commit renames the glo_scalar variables on the interface.

Lots of work has been done to the interface, and that is now reflected in the
documentation.