Speed up the calculation of norms

[UnaiSan]

As it is shown here, there is an improvement in the calculation of norms if instead of

np.apply_along_axis(np.linalg.norm, -1, array)

we write

np.linalg.norm(array, axis=-1)

I have seen this in the following functions/methods: Curve.binormal, Curve.curvature, Curve.torsion, SplineObject.tangent, Surface.normal.

Could this be addressed in a future release?

[VikingScientist]

This is a great suggestion. I was not aware of this discrepancy. Thank you for bringing this to our attention.

[VikingScientist]

Fixed #144