Philippe 19/04/2018: IO: add possibility to choose directory for outputs, backups and diachronic files

* REQ_TAB and T_TX2DP too small if less processes than split files
* missing broadcast if split files (deadlock if IRESP/=0)

Philippe 05/04/2018: IO: set all (except if in TFIELDLIST) SURFEX variables to be not time dependent (temporary?)

Philippe 05/04/2018: IO: modify IO_SET_KNOWNDIMS_NC4, IO_WRITE_COORDVAR_NC4 and IO_WRITE_FIELD_ATTR_NC4 to support files from all programs

This allows to write coordinate variables in all files and with all programs (see also next commits)

needed to manage correctly domain number when writing dimension variables

because they are often needed later (and they do not consume a lot of memory)

Philippe 29/03/2018: problem correction in Rules for gfortran: do not replace AINT by INT (AINT is for reals, INT for integers)
This was the origin of problems with management of XUNDEF in SURFEX
because XUNDEF can not be represented by a standard integer

+ remove temporal_lt function

Philippe 28/03/2018: replace TEMPORAL_DIST by DATETIME_DISTANCE and ADD_FORECAST_TO_DATE and SUBTRACT_TO_DATE by DATETIME_CORRECTDATE

(LTIMEDEP is only set for some test variables)

* set JPXKRK to NLFIMAXCOMMENTLENGTH
* add NSTRINGCHUNKSIZE parameter in mode_netcdf (instead of IMULT)
* typo

Philippe 22/03/2018: IO: use CFORMAT field of TFILEDATA instead of relying on LIOCDF4, LLFIOUT and LLFIREAD
This prevents some problems when mixing LFI and netCDF files (ie in cdf2lfi)
+ cleaning use of modd_io_ll for LIOCDF4, LLFIOUT and LLFIREAD (and others)

* improvements
* split option works again

Philippe 19/03/2018: IO: Z-split files: pass also the optional HFORMAT argument (necessary for LFI2CDF)

Philippe 13/02/2018: IO_GUESS_DIMIDS_NC4: manage variables with CLBTYPE/=NONE + add a return value + problems are now WARNING instead of FATAL