Ensure that chunks is tuple of ints upon array creation

[hanslovsky]

[Description of PR]

This PR will make sure that chunks of an array are created as tuple[int, ..], even if provided as tuple[float, ...] (#1461). This is already the case for shape and this PR ensures that the behavior is the same between those two.

I opened this as a draft PR with a failing test to ensure that github actions fail as expected. I will add the fix and push separately. Please feel free to squash as needed upon merge of the PR.

Fixes #1461

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/tutorial.rst N/A
• Changes documented in docs/release.rst
• GitHub Actions have all passed
• Test coverage is 100% (Codecov passes)

[joshmoore]

(Launched workflows)

[hanslovsky]

Rebased over main

[d-v-b]

normalize_shape already does shape = tuple(int(s) for s in shape), so what is the purpose of decorating this function here?

[hanslovsky]

This is a safeguard for a future additions of early returns. I can remove the decorator here if desired.

[d-v-b]

what about removing the decorator and adding a chunks = tuple(int(s) for s in chunks) line to the end of normalize chunks?

[hanslovsky]

I had that as an option in commit 4693cb8 but there are 3 different return statements in normalize_chunks.

[d-v-b]

in that case I would recommend introducing a _chunks variable in that function that gets returned at the end, and replacing the early returns with assignment to _chunks

[hanslovsky]

Note: None of the early returns in normalize_chunks break my added tests

[hanslovsky]

Also, I am happy with any of these options:

1. decorator on normalize_chunks and normalize_shape
2. decorator on normalize_chunks
3. chunks = tuple(int(s) for s in chunks) at the end of normalize_chunks

I'll pick whatever zarr maintainers think is most appropriate for the code base.

[d-v-b]

I think I prefer option 3, because at the end of the day this is just about ensuring that functions return the correct type, and I don't think we need decorators for that.

[hanslovsky]

I reverted to option 3

[d-v-b]

Looks good!

As an aside, it's weird that we accept booleans for specifying chunks, but I guess it's far too late to change that, unless we were planning a major version increment.

[codecov]

Codecov Report

Merging #1470 (00da620) into main (13cdbff) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main     #1470   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           37        37           
  Lines        14723     14732    +9     
=========================================
+ Hits         14723     14732    +9     

Impacted Files	Coverage Δ
zarr/tests/test_creation.py	100.00% <100.00%> (ø)
zarr/util.py	100.00% <100.00%> (ø)

[joshmoore]

Thanks, guys! Happy to get this in.

Also happy to hear if there is anything that should be done so that mypy could detect that a cast was needed.