Fix memory issue in EBSD refinement due to naive use of dask

[hakonanes]

Description of the change

Fix #468.

I consider this a patch, although a few minor API changes are made:

• EBSD refinement methods accept two optional parameters rechunk, for whether to allow rechunking of pattern array for refinement (which is not the in-place array) and chunk_kwargs, to control the chunk size if rechunk=True.
• When compute=False is passed, EBSD refinement methods return a dask array instead of a list of dask delayed instances. The kikuchipy.indexing.compute_refine_orientation_results() and similar functions accept this array instead of the list.

There is one new issue: the dask array returned when not computing results right away doesn't have the correct shape in its representation, but it produces an array with the correct shape when computed... I'm not sure how to fix this, but decided that it's a too small thing to hold up this patch release.

A 0.5.4 release will be automatically prepared when this PR is merged. I plan to release this on the coming Wednesday.

Progress of the PR

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

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.
• New contributors are added to .all-contributorsrc and the table is regenerated.

[hakonanes]

I'll merge tomorrow and then release v0.5.4.