[OpenVINO] Fix dynamic/symbolic shape handling in get_ov_output and broadcast_to

[goyaladitya05]

Description

When ops.shape() is called on a tensor with dynamic dimensions, it returns a tuple where unknown dims are OpenVINOKerasTensor scalars rather than Python ints. Any downstream op receiving this tuple as a shape argument would fail because get_ov_output's list/tuple branch unconditionally called ov_opset.constant(x), which internally calls np.array() on each element, triggering the symbolic tensor's __array__ guard:

RuntimeError: An OpenVINOKerasTensor is symbolic: it's a placeholder for a shape and a dtype.

My fix for this:
get_ov_output: When the input list/tuple contains any OpenVINOKerasTensor or ov.Output element, build the shape tensor dynamically, convert each element to a rank-1 [1] output (scalar dims are reshaped, static ints become constant([v])), align dtypes, then concat along axis 0. All-static lists continue through the existing ov_opset.constant path.

Contributor Agreement

Please check all boxes below before submitting your PR for review:

• I am a human, and not a bot.
• I will be responsible for responding to review comments in a timely manner.
• I will work with the maintainers to push this PR forward until submission.

[gemini-code-assist]

Code Review

This pull request enhances the OpenVINO backend's get_ov_output function to handle lists containing symbolic tensors, enabling support for dynamic shapes in operations like broadcast_to. Consequently, several dynamic shape tests have been re-enabled. Feedback focuses on improving the robustness of the type checking to include KerasVariable and Tensor, and simplifying the logic by using recursive calls to get_ov_output for individual list elements.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 71.42857% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.95%. Comparing base (d51c980) to head (eac3a3b).
⚠️ Report is 4 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/backend/openvino/core.py	68.42%	3 Missing and 3 partials ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #22692   +/-   ##
=======================================
  Coverage   82.94%   82.95%           
=======================================
  Files         596      596           
  Lines       69200    69272   +72     
  Branches    10806    10821   +15     
=======================================
+ Hits        57399    57465   +66     
- Misses       8973     8976    +3     
- Partials     2828     2831    +3     

Flag	Coverage Δ
keras	82.77% <71.42%> (+<0.01%)	⬆️
keras-jax	58.70% <9.52%> (-0.06%)	⬇️
keras-numpy	54.55% <0.00%> (-0.06%)	⬇️
keras-openvino	59.43% <71.42%> (+0.04%)	⬆️
keras-tensorflow	60.27% <9.52%> (-0.05%)	⬇️
keras-torch	59.04% <9.52%> (-0.07%)	⬇️

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.

[hertschuh]

Thanks for the fix!