feat(python): Enhance bench.py with CLI and default benchmark execution

[Manamama-Gemini-Cloud-AI-01]

This Pull Request enhances the python/scripts/bench.py script by adding a proper command-line interface (CLI) and a default benchmark execution flow.

Background:
Previously, python/scripts/bench.py was designed as a Python module, lacking a if __name__ == "__main__": block for direct execution. This led to a confusing user experience where running the script directly (or with --help) produced no output or executed no benchmark, despite the README.md implying its usability.

Key Changes:

1. Added if __name__ == "__main__": block: The script now includes a standard entry point for direct execution.
2. Integrated argparse: A command-line parser has been added, enabling:
   • python bench.py --help: Displays usage instructions, including options for running benchmarks.
   • python bench.py --run: Executes a quick default benchmark with configurable count, ndim, and connectivity.
   • python bench.py --test_core: Runs a basic functionality test for usearch (Index creation, add, search).
3. Default Benchmark Execution: The --run option now executes a default benchmark using bench_params.
   • Ground truth generation is implemented using usearch.index.search(exact=True) to avoid external dependencies like scikit-learn.
   • The Evaluation object is correctly configured with AddTask and SearchTask to produce numerical results for add_per_second, search_per_second, and recall_at_one.

Benefits:

• Improved Usability: Users can now run bench.py directly from the command line to quickly test usearch functionality and get benchmark figures without writing wrapper scripts.
• Aligned with Documentation: The script's behavior now better aligns with the implicit expectations set by README.md and BENCHMARKS.md.
• Self-Contained Testing: The benchmark now uses usearch's own capabilities for ground truth generation, making it more robust and reducing external dependencies for basic benchmarking.

This PR makes bench.py a more functional and user-friendly tool for quick local testing and performance verification.