Numerical issues in new SAS geometry

When the magnetic field is lowered towards zero in the new SAS geometry, numerical artifacts (2_PolarizedSANS_new.py) appear:

2_new

In contrast, the result computed in 1.18 with the old (1_PolarizedSANS_old.py) geometry (alpha = 0°, instead of alpha = 90°) seems fine and stable:

1_old

In addition, the horizontal centerline is computed differently. Result in new geometry (5_PolarizedSANS_new_center.py):

5_new_center

Compared to result with 1.18 and old geometry (4_PolarizedSANS_old_center.py):

4_old_center

Note: The geometry is still rotated 90° in plane, hence horizontal -> vertical in comparison between 1.18 and current.