add init_array, and data kwarg for create_array

[d-v-b]

Moves much of the logic currently contained in create_array down to a lower-level function called init_array, that creates array metadata + saves it to storage.

By popular request, I added a keyword argument to create_array: data, that allows users to provide an array-like object (default is None), from which dtype and shape can be inferred. I'm not a fan of what this does to the function signature, but at lot of people wanted this 🤷‍♂️ dtype and shape now default to None. If you specify data and dtype, then you get a warning about dtype being ignored in favor of data.dtype, and similarly for shape.

In parallel with #2622, I also added a write_data kwarg that controls whether data is written to the freshly created array. The default value here is True, which IMO is unsafe, but also requested by users.

As an indulgence. I also made some minor changes to the type aliases for ChunkKeyEncoding -- what we were calling ChunkKeyEncodingLike was actually a typeddict that should be ChunkKeyEncodingParams, and ChunkKeyEncodingLike should be ChunkKeyEncoding | ChunkKeyEncodingParams.

closes #2707, and partially addresses concerns raised in #2689

TODO:

• Add unit tests and/or doctests in docstrings
• Add docstrings and API docs for any new/modified user-facing classes and functions
• New/modified features documented in docs/user-guide/*.rst
• Changes documented as a new file in changes/
• GitHub Actions have all passed
• Test coverage is 100% (Codecov passes)

[martindurant]

Thank you

[dstansby]

Nice! I think it would be helpful to not hide away the change to ChunkKeyEncoding here, and put it in it's own PR with it's own bugfix release notes entry. (and that would make the two PRs easier to review too)

[d-v-b]

Nice! I think it would be helpful to not hide away the change to ChunkKeyEncoding here, and put it in it's own PR with it's own bugfix release notes entry. (and that would make the two PRs easier to review too)

good idea, I'll split it out

[d-v-b]

#2763 should go in before this

[martindurant]

The tests look good.
The code diff is quite large for the change - I don't know when I'll have a time to go through it.

[d-v-b]

tests are actually incomplete -- I didn't test the synchronous version of create_array (it doesn't take a data kwarg yet)

[d-v-b]

I will get that in this evening

[d-v-b]

@dstansby any idea why the release notes bot does not believe that I have added release notes? Is there something messed up with my release notes?