Fix fill_value=null zarr v2 to v3 conversion for string dtypes

[observingClouds]

Adds support to convert unset/null fill values from zarr format 2 metadata to format 3 for string types.

• Closes Can't concat virtual datasets without dropping CF variables  #826 (at least the dtype part)
• Tests added
• Tests passing
• Full type hint coverage
• Changes are documented in docs/releases.rst
  - [ ] New functions/methods are listed in api.rst
  - [ ] New functionality has documentation

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.03%. Comparing base (052f7d1) to head (9be6948).
⚠️ Report is 64 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #845      +/-   ##
==========================================
+ Coverage   88.99%   89.03%   +0.04%     
==========================================
  Files          34       34              
  Lines        1945     1943       -2     
==========================================
- Hits         1731     1730       -1     
+ Misses        214      213       -1     

Files with missing lines	Coverage Δ
virtualizarr/manifests/array.py	84.94% <100.00%> (+3.22%)	⬆️
virtualizarr/parsers/dmrpp.py	83.33% <100.00%> (ø)
virtualizarr/parsers/zarr.py	97.46% <100.00%> (-1.29%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[maxrjones]

thanks for working on this, please feel welcome to ping whenever you need a review

[observingClouds]

@maxrjones and others please have a look now. I am not sure if I did the casting correctly to fix the mypy error, but all is passing now. I have also tested that my test is failing with the previous version.

[maxrjones]

Thanks @observingClouds

I think the other fix from #826 (comment) (if not np.issubdtype(self.dtype, dtype):) is also needed to close out #826. Do you want to include that fix in this PR?

[observingClouds]

Alright @maxrjones, I included now the fix mentioned in #826 (comment) and ensured with an additional test that this code is effective (failed without/succeeds with fix)

I also had to do some mypy fix, which I believe is unrelated and originates from a mypy update. Let me know if I shall exclude this from this PR, otherwise I think this is good to go.

[maxrjones]

thanks for this fix!