[OpenVINO] Partial upgrade to opset16 for select operations

[goyaladitya05]

As discussed previously in #22446, introducing a partial upgrade to OpenVINO's opset16, for the ops istft and segment_max. (these are newly added in this opset)

We are not going for a complete upgrade, because there is a bug in opset16 due to some missing ops. It has been fixed upstream, and will be corrected in 2026.1 or 2026.2.

Also refactored the helper fn _segment_reduction_fn into segment_sum, since it is the only caller now.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 5.47945% with 69 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.62%. Comparing base (28a83c5) to head (c4cd8ed).
⚠️ Report is 26 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/backend/openvino/math.py	5.47%	69 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (28a83c5) and HEAD (c4cd8ed). Click for more details.

HEAD has 1 upload less than BASE

Flag	BASE (28a83c5)	HEAD (c4cd8ed)
keras-openvino	1	0

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #22600      +/-   ##
==========================================
- Coverage   83.28%   75.62%   -7.67%     
==========================================
  Files         596      596              
  Lines       68089    68071      -18     
  Branches    10607    10603       -4     
==========================================
- Hits        56711    51480    -5231     
- Misses       8634    14104    +5470     
+ Partials     2744     2487     -257     

Flag	Coverage Δ
keras	75.44% <5.47%> (-7.66%)	⬇️
keras-jax	59.68% <5.47%> (+0.01%)	⬆️
keras-numpy	55.36% <0.00%> (+0.01%)	⬆️
keras-openvino	?
keras-tensorflow	61.05% <5.47%> (+0.01%)	⬆️
keras-torch	59.86% <5.47%> (+0.01%)	⬆️

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.

[gemini-code-assist]

Code Review

This pull request refactors the segment_max and istft functions in the Keras OpenVINO backend to leverage native OpenVINO opset16 operations. The new segment_max implementation needs to be adjusted to correctly handle negative segment_ids and infer num_segments to ensure parity with Keras semantics. Additionally, the istft implementation may require restoring batch dimension flattening logic to properly handle inputs with multiple batch dimensions, as its removal could lead to incorrect results.

[hertschuh]

It looks like something is not working right with ov_istft16 (tests are failing).