Porting DA HOMME to C++

Fine grain todo:

• Rework dirk calling interface so code compiles
• verify eos does not pass
• Determine ColumnOps call to calculate phi_i

Coarse-grained todo:

• Ensure unit tests failing for DA HOMME
• get EOS unit test working.
• Refactor scan_dphi to remove potential duplicate work.
• Get caar_ut unit test to pass (CAAR seems easier?)
• get dirk unit test to pass (dirk seems easier?)

Important notes

• During fortran Dirk routine, phi_i and dphi become desynced. As such, I'm currently assuming that it's best to add a scan to compute_pnh_and_exner in order to calculate phi_i.