spec_driver_fits fails with exit code 11 randomly

[teuben]

spec_driver_fits is called by grid_data.py, but just changing the x_extent/y_extent can cause it to crash.

[teuben]

it's random: re-executing the program it will work again after a few tries. I don't know of a fool proof way to make it crash.

[teuben]

there was an indexing out of the C.caxis[X_AXIS][ix+ii] and C.caxis[Y_AXIS][iy+jj], which was fixed.
the problem should be solved now.

[teuben]

alas, trying out some cases with different cell= and otf_select= cause more of these. electricfence didn't jhelp, it first ran into libnetcdf access with 0 memory allocation problems. frustrating.

[teuben]

non=square maps tend to give predictable segfaults

[teuben]

indeed, axis naming order. no wonder, since X=1 and Y=0 and Z=2..
Sadly the mdarray access method is now 8x slower, possibly because most access is in Z.

[teuben]

perhaps with #5 this is solved. watch out if rmax is large for small test/beam grids with the -a flag. there may still be a crash awaiting resolution.