Docs: Added PyTorch-SU2 coupling example to Python wrapper

[ayush4874]

Closes #2637

Description

This PR adds a comprehensive example (SU2_PY/examples/hybrid_ml_coupling/) demonstrating how to establish a bidirectional coupling between the SU2 solver and an external Machine Learning library (PyTorch) using the Python Wrapper.

Motivation

While SU2 supports data-driven features via MLPCpp, there are currently no explicit examples in the repository showing users how to orchestrate a hybrid "Physics + ML" loop where the solver runs alongside a neural network in the same MPI process. This script bridges that gap.

Implementation Details

The example train_online.py:

1. Initializes a CSinglezoneDriver with a valid mpi4py communicator.
2. Executes a time-stepping loop where the solver advances the physics state.
3. Extracts flow metrics (e.g., RMS_DENSITY) directly from memory in real-time.
4. Performs an online training step on a PyTorch surrogate model using the live simulation data.

Verification

• Tested on Linux (WSL2/Ubuntu 24.04).
• Validated with SU2 compiled using -Denable-pywrapper=true -Dwith-mpi=enabled.
• Confirmed simultaneous execution of C++ solver and Python training loop.

This serves as a starting point for users exploring Physics-Informed Machine Learning (PIML) or online surrogate modeling with SU2.

PR Checklist

• I am submitting my contribution to the develop branch.
• My contribution generates no new compiler warnings.
• My contribution is commented and consistent with SU2 style.
• I have updated appropriate documentation (Tutorials, Docs Page, config_template.cpp), if necessary.

[bigfooted]

I think this example is best placed in the tutorials:
https://github.com/su2code/Tutorials
and you can then add a page to the website repository in the tutorials section:
https://github.com/su2code/su2code.github.io
Best is if it is also regression-tested so the tutorials do not break or get out of date when solvers are updated.

[ayush4874]

Thanks for the feedback @bigfooted. I agree that the Tutorials repository is a better home for this example, as it serves more as a workflow demonstration than a core codebase change.

I will close this PR and proceed with the following plan:

1. Migrate Code: Submit the example script and configuration files to a new PR in su2code/Tutorials.
2. Add Documentation: Submit a corresponding documentation page to su2code/su2code.github.io to explain the hybrid coupling workflow.

Regarding regression testing:
I would love to add this as a regression test to ensure the Python wrapper capabilities remain stable. However, this specific example requires torch. Does the current CI environment include PyTorch, or should I structure the test to skip gracefully if the dependency is missing?