Fix module_mp_radar USE dependency invisible to CMake scanner

[mathomp4]

I had Claude try and figure this out. This was a thought it had...

---

Problem

GCC and ifort fail to compile several files due to standard violations that ifort accepted silently as non-standard extensions.

Fix 1 — Process_Library.F90: hoist USE module_mp_radar to module level

GCC: Fatal Error: Cannot open module file 'module_mp_radar.mod' for reading
ifort: error #6404: This name does not have a type [XAM_R]

Root cause: CMake's Fortran module dependency scanner only reads USE statements at the module/program specification part (top level). It does not see USE statements inside contained subprograms.

Process_Library.F90 had USE module_mp_radar inside two contained subroutines (init_refl10cm and calc_refl10cm), making the dependency on module_mp_radar.F90 invisible to CMake. This was a latent bug that never fired before because compilation order happened to put module_mp_radar.F90 first by coincidence.

Commit 37878b71 changed aer_actv_single_moment.F90 to USE GEOSmoist_Process_Library, which introduced a new intra-library dependency that reshuffled the compilation schedule and exposed the hidden gap.

Moving USE module_mp_radar to the module-level USE block of GEOSmoist_Process_Library makes the dependency visible to CMake. The symbols are then available to init_refl10cm and calc_refl10cm via host association, which is semantically identical.

Fix 2 — GEOS_MoistGridComp.F90: replace not() with .not.

GCC: Error: 'i' argument of 'not' intrinsic at (1) must be INTEGER

Root cause: not() is the bitwise integer intrinsic; .not. is the logical negation operator. ifort silently accepts not() on logicals as a non-standard extension, but GCC correctly rejects it.

! Before
if ((USE_AEROSOL_NN) .and. not (USE_NCLOUD_CLIM)) then
! After
if ((USE_AEROSOL_NN) .and. .not. (USE_NCLOUD_CLIM)) then

Fix 3 — GEOS_PhysicsGridComp.F90: pad strings in CHARACTER array constructor

GCC: Error: Different CHARACTER lengths (3/2) in array constructor

Root cause: The Fortran standard requires all elements of an array constructor to have the same character length. 'KM' and 'RI' (length 2) were mixed with 'ALH' (length 3). ifort accepted this silently; GCC correctly rejects it.

! Before
SHORT_NAME  = (/'ALH', 'KM', 'RI'/),
! After
SHORT_NAME  = (/'ALH', 'KM ', 'RI '/),