Move from ESMA_cmake to f2py-cmake

[mathomp4]

Per @amdasilva, QFED (and GMAOpyobs) are pure Python compare to a lot of GEOS-ESM packages. As such, they really have no need for ESMA_cmake or ecbuild, say. However, they do use f2py a lot and our f2py CMake code is in ESMA_cmake.

But, @amdasilva pointed me to f2py-cmake by @henryiii.

I have a first rough version of the transition in the feature/mathomp4/plume-no-esma branch (which is based off of feature/#17/adasilva/plume_rise)

ETA: NOTE: The f2py-cmake code in my branch is a modified version of the stock code, see scikit-build/f2py-cmake#21 where we have some fixes to allow it to be run more than once.

[mathomp4]

As noted by @amdasilva when he tested this, this does build but it doesn't run.

For example:

> python3 -c 'import FreitasPlume_'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
ImportError: /gpfsm/dswdev/mathomp4/Models/QFED-Arlindo-f2pycmake/install/lib/Python/plumerise/FreitasPlume_.cpython-311-x86_64-linux-gnu.so: undefined symbol: plume_

and the same for LockePlume_:

> python3 -c 'import LockePlume_'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
ImportError: /gpfsm/dswdev/mathomp4/Models/QFED-Arlindo-f2pycmake/install/lib/Python/plumerise/LockePlume_.cpython-311-x86_64-linux-gnu.so: undefined symbol: plumeg_

[mathomp4]

I thought I had one fix which is that UseF2Py.cmake does not realize that our .F90 files are Fortran 90. The code has:

    foreach(file IN LISTS ALL_FILES)
        if("${file}" MATCHES "\\.f90$")
            set(HAS_F90_FILE TRUE)
            break()
        endif()
    endforeach()

this is simple enough to fix by adding a:

        if("${file}" MATCHES "\\.F90$")
            set(HAS_F90_FILE TRUE)
            break()
        endif()

in there.

But this doesn't seem to do much.

[mathomp4]

One more thing I notice is that the f2pywrapper files are empty:

> \ls -l *wrap*
-rw-r--r-- 1 mathomp4 g0620 0 Dec 17 09:14 FreitasPlume_-f2pywrappers.f
-rw-r--r-- 1 mathomp4 g0620 0 Dec 17 09:14 FreitasPlume_-f2pywrappers2.f90
-rw-r--r-- 1 mathomp4 g0620 0 Dec 17 09:14 LockePlume_-f2pywrappers.f
-rw-r--r-- 1 mathomp4 g0620 0 Dec 17 09:14 LockePlume_-f2pywrappers2.f90

So it's almost like the code isn't actually doing what I think it should. My call sequence is pretty boring:

# Library: plumerise
# ------------------
add_library (plumerise
  LockePlume_Mod.F90
  rconstants.F90
  FreitasPlume_Mod.F90
  qsat_Mod.F90
)
install(TARGETS plumerise DESTINATION lib/Python/plumerise)

# Module: FreitasPlume
# --------------------
f2py_object_library(FreitasPlume_object OBJECT)
f2py_generate_module(FreitasPlume_ FreitasPlume_py.F90 OUTPUT_VARIABLE FreitasPlume_files)
python_add_library(FreitasPlume_ MODULE "${FreitasPlume_files}" WITH_SOABI)
target_link_libraries(FreitasPlume_ PRIVATE Python::NumPy)
target_link_libraries(FreitasPlume_ PRIVATE FreitasPlume_object)
target_link_libraries(FreitasPlume_ PRIVATE plumerise)
target_link_libraries(FreitasPlume_ PRIVATE OpenMP::OpenMP_Fortran)
install(TARGETS FreitasPlume_ DESTINATION lib/Python/plumerise)