LMDZ.MARS + MESOSCALE
A quite major commit, at least for MESOSCALE.
In a word: any ngrid deserves to be free.
- no need to recompile when changing number of horizontal grid points or number of processors
- latest version of LMDZ.MARS physics can be used
- WARNING! Nesting is still yet to be fixed (since r1027)
Also some small bug fixes to LMDZ.MARS.
Changes in LMDZ.MARS
--> fixed a potential bug in thermal plume model because zlmax was computed both in thermcell_main_mars and calltherm_interface... so made it an OUT argument of calltherm_interface. also: changed the name to limz. and added precompiling flags to avoid the use of planetwide in MESOSCALE. in MESOSCALE we just go high enough (nlayer-5) and do not care about computational cost (although we certainly gain from not using MAXVAL).
--> moved allocations upward in inifis. does not change anything for GCM, but make MESOSCALE modifications simpler, and overall make inifis better organized: first allocations, then reading callphys.def file.
--> added precompiling flags around lines that are both useless for MESOSCALE (notably I/O) and recently adapted to parallel computations in the GCM
--> tidied up what is MESOSCALE vs. GCM in surfini
Changes in MESOSCALE
--> changed makemeso to allow dynamically set nx ny nprocs
--> changed makemeso to remove links to Fortran code adapted to parallel GCM and useless for mesoscale
--> changed ngridmx to ngrid in inifis includes
