- The minimum viscsoity was only being applied in the strain component.
  The minimum with KH_BG_MIN was not applied to the tension component.
  This broke isotropy of the stress tensor.
- The parameter KH_BG_MIN defaults to 0.0 which would be harmless if
  viscosities are always non-negative. However, when using backscatter
  the viscosity is intentionally negative and this limiter was zeroing
  out the backscatter, but only in the strain component.
  - I've moved the min() function to before adding in the MEKE component
    which is where negative viscosities might come from.
- This does not change answers in any tests BUT will change answers for
  any backscatter experiments OR where KH_BG_MIN is non-zero.

- Delete unused supercritical bathymetry.

  Added a new pointer, Kv_shear_Bu, to the vertvisc_type to describe the
viscosity at the corner points. Also amended the comments describing TKE_turb
to note that this now can be discretized at either the tracer or corner points.
By default all answers are bitwise identical.

  Check whether visc%Kv_shear_Bu is allocated, and if so add it to the
viscosity.  All answers are bitwise identical unless visc%Kv_shear_Bu is
allocated.

  Added a new subroutine, Calc_kappa_shear_vertex, to calculate the Jackson et
al. stratified shear mixing at the cell corners, along with a new runtime
parameter, VERTEX_SHEAR, to control wheter this is used, and a second new
function, kappa_shear_at_vertex, to allow external queries of the setting of
VERTEX_SHEAR.  By default, all answers are bitwise identical, but there are
two new interfaces and MOM_parameter_doc files have a new field.

  Added code to calculate shear mixing at the cell verticies if VERTEX_SHEAR is
true.  Also added calls to allocate visc%Kv_visc_Bu and visc%TKE_turb in this
case, including the addition of new restart fields.  By default, all answers are
bitwise identical, but several recently added interfaces are used.  With this
commit, the vertex mixing is fully available.

  Added halo updates of vertex turbulence fields during initialization for
reproducibility across restarts in non-symmetric memory mode.  Also eliminated
an unnecessary variable.  All answers in test cases are bitwise identical.

Anisotropic horizontal viscosity

…into adcroft-bugfix-minimum-visc

  Corrected a typo in the restart registration call for Kv_shear that was
introduced with Hallberg-NOAA/MOM6@130a32b.  This bug fix should restore the
ability to reproduce across restarts.

Add vertex shear mixing

- turning on COMPUTED_VORTICITY makes things worse at outflow end.

  Eliminated the file MOM_OCMIP2_CO2calc.F90, which is not being called by any
MOM6 code.  This file violates numerous MOM6 coding and licensing conventions,
including some that break the automatic generation of documentation by MOM6, and
it is an unused duplicate of capabilities that are found in the generic tracer
code. All answers in test cases are bitwise identical.

- Split single declarations of multiple variables which leads to
  undocumented parameters.

- Split single declarations of multiple variables which leads to
  undocumented parameters.

- Doxygen comments in MOM_memory.h have to be specially formatted because
  of the ".h" extension (making it behave C-like) and it tooks some
  experimentation to find the one format that allows a comment to be split
  across multiple lines
- Doxumentation of CPP macros only works for #define so the #undef lines
  have to be moved to the end.
- Also reformatted config_src/dynamic/MOM_memory.h to look like a copy
  of config_src/dynamic_symmetric/MOM_memory.h, except for the one macro.

- The documentation for MOM_ice_shelf_dynamics said it was the
  thermodynamics - corrected.
- All of the module doxumentatino was a complete copy of the
  MOM_ice_shelf module which caused conflicts with section and names
  and was incorrectly duplicated. This has been cut.

- Added section labels

- coord_adapt.F90, coord_rho.F90 and coord_slight.F90 each had single
  declaration lines for multiple variables leading to undocumented
  variables.

- Split single declarations of multiple variables which leads to
  undocumented parameters.

- Split single declarations of multiple variables which leads to
  undocumented parameters.

- Document a variable (e.g. Time) with the comment "!< Time" apparently
  is an error. Admittedly the comment gives no extra information than
  the variable name. I've changed the comment to "!< Model time".

Conflicts:
	src/framework/MOM_restart.F90

Merge candidate 2018 07 16

Conflicts:
	src/core/MOM.F90
	src/core/MOM_variables.F90

Add melt/freeze potential via MCT cap

add latitudinally varying background diffusivity

Improve melt potential calculation

Makes OBLD available to MCT (and other caps)