fix: allow opening non-existent ZipStore in read mode

[NIK-TIGER-BILL]

Summary

Currently, opening a non-existent ZipStore in mode r fails because zipfile.ZipFile raises an exception when the file does not exist. This is inconsistent with other Store classes which can be opened in r mode without issues.

Changes

• Handle the case where a ZipStore is opened in r mode but the file does not yet exist.

Testing

• Added test verifying that a fresh ZipStore can be opened in r mode.

Fixes #2450

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.53%. Comparing base (b9d3964) to head (93b76fd).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4048   +/-   ##
=======================================
  Coverage   93.53%   93.53%           
=======================================
  Files          88       88           
  Lines       11894    11897    +3     
=======================================
+ Hits        11125    11128    +3     
  Misses        769      769           

Files with missing lines	Coverage Δ
src/zarr/storage/_zip.py	97.72% <100.00%> (+0.03%)	⬆️

🚀 New features to boost your workflow:

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

[d-v-b]

i feel like it's a bit surprising if mode=r has side effects (creating a zip file). This is a special thing for the zipstore because a zip archive exists in some other storage system.

[NIK-TIGER-BILL]

That is a fair point. I chose creating an empty zip as the least-surprising way to keep the store API consistent — other stores can be opened in r mode even when the backing path does not exist yet, and the caller gets an empty read-only store.

The alternative I considered was to raise a custom FileNotFoundError, but that breaks the uniformity of the Store.open contract. Another option would be to hold a no-op ZipFile-like object internally, but that would require re-implementing parts of zipfile just to avoid the side effect.

I'm happy to switch to a different approach if the maintainers prefer — e.g., raise a clear FileNotFoundError in r mode, or document the side effect explicitly. Let me know what you think.