+Refactor spatial means and rescale some global integrals

[Hallberg-NOAA]

This PR consists of two commits that use the unscale arguments to reproducing_sums() to refactor the 6 global area mean functions (global_area_mean(), global_area_mean_u(), global_area_mean_v(), global_layer_mean(), global_volume_mean() and adjust_area_mean_to_zero()) and to revise the scaling of the returned values from the two global integral
routines(Global_area_integral() and global_mass_integral()), and then to use this revised capability to work exclusively with rescaled diagnostics of the surface fluxes.

When global_area_integral() and global_mass_integral() are called with a tmp_scale argument, now it is not just the variable that is being integrated that is returned in rescaled units, but also the area or mass, whereas previously it was only the variable that was returned in rescaled units but the area or mass multiplying them were in mks units of [m2] or [kg]. In other words, these routines now return variables in units of [L2 A ~> m2 a] and [R Z L2 A ~> kg a] (if tmp_scale is present) or [m2 a] and [kg a] (if it is not), but no longer in mixed units of [m2 A ~> m2 a] and [kg A ~> kg a]. As a result the code surrounding the 4 instances where global_area_integral or global_mass_integral were being called with tmp_scale arguments (in MOM.F90 and MOM_ice_shelf.F90) also had to be modified in the same commit.

This same commit also includes a rescaling in the units of the areaT_global and IareaT_global elements of the ocean_grid_type and dyn_horgrid_type to [L2 ~> m2] and [L-2 ~> m-2], respectively. These elements were only used in the area_mean functions, so it is sensible to make this change in the same commit where the area_mean functions were revised. Although the dyn_horgrid_type is shared between MOM6 and SIS2, these elements are not used in SIS2.

The second commit in this PR keeps 36 area integrated surface mass, heat or salt flux diagnostics in forcing_diagnostics() in rescaled units by replacing an unscale argument with a tmp_scale argument to global_area_integral(). It also adds the corresponding conversion arguments to the register_scalar_field() calls for each of these diagnostics, so that the documented units and conversion factors can be used to confirm the correctness of the documented units for these variables.

All answers and diagnostics are bitwise identical, but there are changes in the rescaled units of two elements each in two transparent types, and changes to the rescaling behavior of two publicly visible routines when they are called with tmp_scale arguments. This commits in this PR include:

• b3ed3f541 Diagnose area integrated fluxes in rescaled units
• 4e0c742de +Refactor the spatial mean calculations