Several tendencies in marbl_ciso_mod.F90 are missing zoo_graze terms; the first
step to add them in is to make them accessible in the module via the share
type.

Previously, marbl_ciso_mod.F90 computed tendencies without considering
the possibility that zooplankton could be grazed, so no zoo_graze_*
terms were included. The mass balance check did not catch this (even in
a test run where zoo grazes itself) because the way the mass balance
checks are constructed the missing terms all canceled each other out.

I did confirm that adding some zoo_graze terms (but not all of them)
triggered the error when zoo grazed itself and had no effect when zoo
was not grazed.

marbl_ciso_mod.F90 only wants the sum of zooplankton terms like zoo_loss
and zoo_graze (sum over all zooplankton species). Instead of sending
each individual loss / grazing term, marbl_mod.F90 now does the sum in
marbl_export_zooplankton_shared_variables().

Also, marbl_ciso_mod.F90 does not need x_graze_zoo so that field is not
included in the share type.

Instead of copying these into marbl_autotroph_share_type, just pass
autotroph_secondary_species to marbl_ciso_set_interior_forcing().

Cleaned up more variables that were available to marbl_ciso_mod.F90 via
autotroph_secondary_species but were still being copied to
marbl_autotroph_share_type. At this point, the only use of the share
type is to copy tracer_local terms and there is already a plan to pass
tracer_local instead.

Part of some post-review code-cleanup - realized that the
zooplankton_share_type and marbl_ciso_mod.F90 should be using zootot instead of
zoo to signify this is a sum over zooplankton species. As a first pass, I
changed zoo13Ctot and zoo14Ctot to zootot13C and zootot14C, respectively.

Since marbl_mod.F90 sums over all zooplankton when storing data in the
zooplankton share type, those variables are now refered to as zootot:

zoototC
zootot_loss
zootot_graze

And so on. This is most pronounced in marbl_ciso_mod.F90, but also is apparent
in the diagnostics module.

The changes to diagnostic names and metadata from the previous commit
were not included in default_diagnostics.yaml, so GCMs were looking for
CISO_zooC_d1[34]C instead of CISO_zoototC_d1[34]C

Expand autotroph_local_type to include carbon isotope elements, and then pass
autotroph_local to marbl_ciso_set_interior() instead of copying data into the
share type. Other related clean-ups:

* marbl_autotroph_share_type is no longer needed
* marbl_compute_autotroph_phyto_diatoms() does not need autotroph_local
* marbl_ciso_mod.F90 no longer needs its own autotroph_local_type, so there is
  no longer a consistency check there (or a setup_autotroph_local, since that
  is taken care of in marbl_mod)

tracer_local is no longer restricted to the base tracers, it also
contains carbon isotope tracers if ciso_on = .true. This fixes a bug in
attempts to put carbon isotope fields in autotroph_local.

No need to compute this ratio for autotrophs that do not have CaCO3
tracer

Instead of copying some elements of tracer_local to interior_share and
recomputing other elements in setup_local_column_tracers, just pass
tracer_local (instead of tracers) to marbl_ciso_set_interior_forcing(). This
means that some associate statements need to be updated to point to elements of
tracer_local instead of marbl_interior_share_type (and some local variables are
now associate statements instead of being computed in marbl_ciso_mod).

DO13Ctot_loc was inadvertently pointing to DO14Ctot_loc

* several "zooC" -> "total zooC" or "zoototC" in comments
* R1[34]C_zooC -> R1[34]C_zoototC
* autotroph consistency check has a single "if (ciso_on) then" block
  and checks for Ca1[34]CO3_ind instead of individual checks for all
  isotopic autotroph tracers
* ciso_set_interior_forcing(): reordered arguments in interface so all
  intent(in)s are clumped together, followed by the (inout)s