Wrong scaling of snow depth and convective precip when reading ECMWF input in GRIB2

[pirmink]

Flexpart expects precipitation in mm/h and snow depth in m water equivalent. This is the same for GRIB1 and GRIB2.

For GRIB2 however, when the input field is SD or CP, the conversion_factor=1000. is set in readwind_ecmwf.f90. Later in this subroutine, the values of the SD and CP fields read from GRIB2 are divided by conversion_factor, i.e. by 1000.

This seems to be wrong, as flex_extract outputs precipitation in mm/h, which is what flexpart expects according to com_mod.f90 and get_wetscav.f90. SD is in m, as expected by flexpart according to getvdep.f90. In both cases no conversion is needed in flexpart.