Adapt LSTATNW to GPU
move the key declaration from modd_cturb to modn_turbn and add the key
as args in shallow + turb

In the previsous commit the physics modifications between cy48t1 and cy48t3
have been introduced in PHYEX but still using the cy48t1 pack.
This commit introduces new modifications (in arome/ext) to use the cy48t3 pack.
Results are different due to modifications introduced betwwen cy48t1 and cy48t3
in parts other than the physics.
So, a new reference is needed (modifications in check_commit_ial.sh, INSTALL_pack_ial.md).

Modifications needed after the cleaning done by Ryad

Quentin 19/08/2022: bugfix at merge (mode_prandtl AROME version was used instead of MNH for RED variables)

Quentin 19/08/2022: packing tke_eps_sources uniformize use of IIJB:IIJE + correct bug on vertical loop index

Quentin 18/08/2022: move 3D computation of rmc01 (optional part) in rmc01_3D in case RMC01 tested in AROME (with horizontal packing)

Quentin 18/08/2022: Incomplete NPROMA bloc test lead to correction on loop index computation (IIJB:IIJE)

Quentin 16/08/2022: Prepare hpacking of turb.F90 (END) + internal subroutines (DELT,DEAR,COMPUTE_FUNCTION_THERMO, CLOUD_MODIF_LM)

- Mesonh: add ext/ibm_affectv and remove USE MODI_IBM_MIXINGLENGTH (not used)
- AROME : mnh_expand in tm06 optimization and simplify ibm_mixinglength

- source_neg_correct : Explicit dim. declaration + version for AROME/MESONH different (for Packing)
- ibm_mixinglength : Explicit dim. declaration (3D vars must be kept : horizontal computations ==> turb 3D)
- rotate_wind and update_rotate_wind : Explicit dim. declaration in (3D vars must be kept : horizontal computations ==> turb 3D)
- move turb.F90/update_rotate_wind within mode_rotate_wind.F90
- mode_sbl_phy : Packing LMO + uniformize packing
- mode_tm06 : Packing
- turb.F90 : adapt CALLs

+ add new shuman functions that takes 2D array (i,j) instead of 3D for specific computation in turb_ver_dyn
+ expand last parts of turb_ver_dyn
+ remove USE MODD_OCEANH and add ocean args as intent(in)

TODO: remove index:index by index and add mnh_expand again on turb_ver_dyn

Quentin 12/08/2022: Expand + explicit dims mode_turb_ver_dyn (END) : OCEAN + L2D + prepare for hor. packing

Quentin 12/08/2022: Complete cross gradients and shuman as subroutines with explicit dimensions declarations