fix: use canonicalize_axis function in reduce_shape function

[miguelteixeiragomes]

Pull request to fix issue #21648

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.55%. Comparing base (08b5252) to head (7971d78).
⚠️ Report is 8 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #21650   +/-   ##
=======================================
  Coverage   82.55%   82.55%           
=======================================
  Files         571      571           
  Lines       57618    57646   +28     
  Branches     8999     9008    +9     
=======================================
+ Hits        47564    47589   +25     
- Misses       7759     7760    +1     
- Partials     2295     2297    +2     

Flag	Coverage Δ
keras	82.35% <100.00%> (+<0.01%)	⬆️
keras-jax	63.49% <100.00%> (-0.03%)	⬇️
keras-numpy	57.84% <100.00%> (-0.03%)	⬇️
keras-openvino	34.42% <100.00%> (+0.03%)	⬆️
keras-tensorflow	64.20% <100.00%> (-0.03%)	⬇️
keras-torch	63.71% <100.00%> (-0.03%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[fchollet]

Thanks for the PR! Please add a unit test for the fix.

[miguelteixeiragomes]

@fchollet , sure, added a test with multiple negative axes, but without keepdims. Let me know if you think more tests are needed.
There is a test failing related to quantization, but I don't understand if there's any relation to the test I added.
Never mind this part, after a re-run if passed.
Thank you!