Minor docstring/behaviour mismatch in 2D and 3D coordinates/geometries.

[OceanNuclear]

Describe the papercut

The input to bluemira.geometry.coordinates.normal_vector should be of shape (2, N), not (N, 2).

This is so minor that it'll take about 10 seconds to fix. But I just needed to raise a papercut because I'm currently busy working on a branch that I've not committed.

To turn this into a proper papercut, maybe I'll propose one more thing to do/check:

1. Run grep -r "2, N", grep -r "N, 2", grep -r "3, N", grep -r "N, 3" on bluemira/geometry/ and bluemira/codes/. They should return about 15 instances. My question is: why do some use (2, N)/(3, N) format, while others take the (N, 2)/(N, 3) format? Is there a rhyme or reason, and if not, can we have the same format everywhere? (I guess the papercut can be to turn this into a discussion if we don't come to an agreement? Or simply make the change and commit it if the conclusion is simple and straightforward.)