curvature sign flip

[akhanf]

I haven't been looking for this recently, but I recall it was an issue when we moved to new surface generation (e.g. I think the surface normals can point in opposite directions for different surfaces).. This is a reminder to see if this occurs in multihist7, before we finalize our atlas..

[jordandekraker]

So far in looking at my wetrun testing the curvature seems to be in the correct direction, but we should definitely keep an eye out

Also curvature does seem to still include a few extreme values, even though we are running on smoothed surfaces. We could consider bringing back np.tanh() to normalize this, or just np.clip()