Philippe 11/12/2023: OpenACC: contrav: add condition for data present for some variables (available or not depending on the condition)

Juan 22/11/2023:modeln.f90, Bypass Bug with nvhpc/22.11 <-> cuMemcpyDtoHAsync returned error 1 => move ( NSV /= 0 ) test out of kernels

Juan 14/11/2023:init_mnh.f90, add initialisation of GPU halo with INIT_HALO_D , to avoid some potentiel null pointer a first GET_HALO_D call

Juan 14/11/2023: rain_ice_fast_ri.f90, Bug/data not present with CCE/16.0.1 around where -> add mnh_expand_where to bypass the bug (other where have already been transformed this way)

Juan 27/10/2023:prep_ideal_case.f90, for OPENACC add initialisation of MNH_ALLOC_ZWORK+INIT_HALO_D => PREP_IDEAL work with VER_USER=ZSOLVER version

Juan 24/10/2023:MNH/*.f90 , for cce/16.0.0 , add new present_cr directive to avoid FALSE RECURRENCE on GPU

Philippe 10/10/2023: tridz: remove wrong condition (was there to prevent problem with ZSOLVER; correct condition is based on HPRESOPT)

Philippe 26/09/2023: ini_modeln: allocate XBFB to zero size if not used (useful to prevent crash with OpenACC)

Philippe 19/09/2023: OpenACC: PREP_IDEAL_CASE: put some data on GPU (remark: not expected to work on GPU yet)

Philippe 18/09/2023: FLAT_INV: fix for NOOPENACC (deallocate ZBAND_YT and ZBAND_YRT only if allocated)

Philippe 12/09/2023: remove key MNH_PGI (not useful anymore, was to workaround an bug in very old version of PGI compiler)

Philippe 29/08/2023: OpenACC: work around for a problem with NVHPC 23.7 (not present in older versions)

(cherry picked from commit ce39d7dc)

Philippe 29/08/2023: OpenACC: work around for a problem with NVHPC 23.7 (not present in older versions)

Philippe 25/07/2023: transfer ZSOLVER/ppm.f90 to MNH/ + adapt it. This gives better performance on GPUs.

Philippe 17/07/2023: FFT: invert loops for better performance and to allow correct parallelization by NVHPC OpenACC

Philippe 26/06/2023: ice4_sedimentation_split: remove a DO CONCURRENT because of a reduction in loop (wrong reults with NVHPC 23.5)

(cherry picked from commit ad799fd0)

(cherry picked from commit 18a466fd)

Philippe 26/06/2023: ice4_sedimentation_split: remove a DO CONCURRENT because of a reduction in loop (wrong reults with NVHPC 23.5)