Bug fix in fxhyp. There was some bad logic for the shifting of
longitude grids rlonuv, rlonv, rlonm025 and rlonp025 toward [- pi,
pi]. In some cases, one of the four grids was not shifted and the
others were. For example, you could see the bug with: iim = 48, clon =
20, grossismx = 3, dzoomx = 0.15. The bad logic involved the variable
is2 in the loop on ik = 1, 4. The loop included some tests depending
on ik. The whole thing was quite convoluted. Rewrote fxhyp. In
particular, replaced the loop on ik by a call to a new procedure,
invert_zoom_x. fxhyp.F was in dyn3d, it is replaced by fxhyp_m.F90
which is in dyn3d_common. Removed several arguments of fxhyp: zoom
parameters are accessed through serre.h; rlonm025 and rlonp025 were
not used in inigeom; min and max of longitude steps are written inside
fxhyp.
Some simplifications and modernizations in fyhyp. In particular,
several arguments are removed: zoom parameters are accessed through
serre.h; yprimv was not used in inigeom; min and max of latitude steps
are written inside fyhyp.
Removed now useless intermediary procedure fxyhyper.
Changed default value of dzoomx and dzoomy from 0 to 0.2. dzoom[xy] is
only needed when grossism[xy] > 1 and then it should be > 0.
dzoom[xy] can no longer be the extent of the zoom in degrees: it must
be expressed as the fraction of the total domain.
