[WIP] Evaluation of platinum bench locally

[Vahe1994]

This PR makes several changes to Platinum Bench:

1. Platinum Bench is now an installable package.
2. Adds support for local evaluation using vLLM.
3. Adds a command-line interface (CLI).
4. Enables usage from external projects.
5. Adds Hugging Face model support.
6. Integrates W&B logging (including error artifacts).
7. Improves seed handling.
8. Updates the README and documentation.

• Add vLLM support to evaluate models locally
• Make platinumbench an installable Python package
• Create a CLI and make it runnable from the terminal
• Add an interface so other projects can call platinumbench
• Add support for Hugging Face models
• Add W&B logging, including error artifacts
• Correct and update documentation
• Test that local integration works
• Fix random seed handling
• Add the seed to the run key/identifier
• Test with HSUW
• Enable reasoning for Qwen models
• Verify that OpenAI, vLLM, and local generation produce consistent results
• CLI testing with vLLM serving
• Add local evals for VLM
• Check that evaluation with online API requests still work for LLM
• Test correctness
• Update the README after the changes
• Clean up the code
• Add and fix project documentation
• Check that evaluation with that online API requests still work for Visual VLM