Model sends wrong tile_num to CCPP for nests

[SamuelTrahanNOAA]

Description

In a global simulation with a static nest, the nest thinks it is on tile 1. Some suites read per-tile static data. The nest data must come from tile 7, but the nest tries to read tile 1 instead. It happens because CCPP receives tile_in_mosaic instead of global_tile from the atmos_cubed_sphere.

To fix this, we'll need to get the global_tile from the dynamical core. This issue documents that:

• fv3atm needs the global_tile from atmosphere_control_data so CCPP can read per-tile static files NOAA-GFDL/GFDL_atmos_cubed_sphere#329

To Reproduce:

What compilers/machines are you seeing this with?
N/A. Machine and compiler don't affect it.

Give explicit steps to reproduce the behavior.

1. Run a global simulation with a static nest.
2. Print the tile_num, or use a suite that reads in per-tile static data
3. Tile_num is wrong (or model crashes due to reading the wrong tile).

Additional context

I have a fix for this already, which I'll PR soon. It includes a fix for another bug (or two) which I must fix to run the test case.

Testing:

Have you tested the code changes?

Yes. I have a test case of the GFS with a static nest using current GFS physics. It has per-tile static data for all seven tiles.

On what platforms?

Hera Intel Centos.

Have you run regression test in ufs-weather-model or ufs-s2s-model with code changes?

Not yet.

Will the baseline results change?

Will know that after I run the regression tests.