Description of the change

ax2qu conversion function throws if the arrays are multidimensional and last axis size is 1, even though they are broadcastable.

I appreciate having a redundant axis in an array is not ideal, but sometimes it is easier to work with it to keep the input and output shapes consistent 😄.

I have looked into the code and there appears to be a lot of shape juggling because the internal functions are numba jitted and ultimately require 2d arrays. These lines in particluar seem to be the issue:

I believe we should also check here if axes.shape[:-1] == angles.shape and this will fix the issue.

Progress of the PR

• [-] Docstrings for all functions
• Unit tests with pytest for all lines
• Clean code style by running black via pre-commit

Minimal example of the bug fix or new feature

>>> from orix.quaternion import Quaternion
>>> import numpy as np

>>> axes = np.random.randn(2, 1, 3)
>>> angles = np.random.randn(*axes.shape[:-1])

>>> Quaternion.from_axes_angles(axes, angles)
ValueError: The dimensions of axes (2, 1) and angles (2,) are incompatible. The dimensions must match or one must be a singular value.

With the fix in this PR:

>>> Quaternion.from_axes_angles(axes, angles)
Quaternion (2, 1)
[[[ 0.9999 -0.0058 -0.0025 -0.0078]]

 [[ 0.9386 -0.1434 -0.1071  0.295 ]]]

For reviewers

• The PR title is short, concise, and will make sense 1 year later.
• New functions are imported in corresponding __init__.py.
• New features, API changes, and deprecations are mentioned in the unreleased
  section in CHANGELOG.rst.
• Contributor(s) are listed correctly in __credits__ in orix/__init__.py and in
  .zenodo.json.