Necessary changes for land implicit coupling

[JosephMouallem]

These changes are necessary for implicit land coupling.
1- dkt (heat diffusivity) is now saved in statemid to be used in the surf_diff calculations
2- The tridiagonal matrix is flipped upside down and is now solved from the top of the atmosphere down to the surface and back up
3- Added necessary routines, mostly taken from the AMx physics, to atmos_shared for an easier SHiELD & AMx merge.

surf_diff is populated in NOAA-GFDL/GFDL_atmos_cubed_sphere#408 which is called from NOAA-GFDL/atmos_drivers#71

[lharris4]

The file vert_diff.tech.psis hard to find in more recent code distributions. I have attached a PDF conversion of this file with this comment, which explains the nomenclature in populate_surf_diff() and the routines in this module.
vert_diff.tech.pdf