Add C++ linear SVM training example with automatic differentiation

[StarxSky]

• ☑️ I understand it is strictly prohibited to use AI to write PR description

Summary

This pull request adds a C++ example demonstrating how to
implement and train a linear Support Vector Machine (SVM)
using MLX's automatic differentiation capabilities.

It implements a binary classifier using hinge loss
with L2 regularization and gradient-based optimization.
So, this code can be used to learn how to build machine learning algorithms using mlx.

The code has passed the CLI check.

Testing

I have tested the example locally using :

• Operating System: macOS Tahoe 26.0.1
• Architecture: arm64 M1 Processor
• Compiler: [Apple clang version 17.0.0 (clang-1700.6.3.2)]
• MLX version: [MLX 0.32.3]