feat: add default option using transformers

[EYH0602]

No description provided.

[Copilot]

Pull Request Overview

This PR adds support for using Hugging Face Transformers models as a default fallback option when no specific model provider is matched. The change removes the VLLM dependency and replaces it with a Hugging Face implementation.

• Adds new HFChat class using Hugging Face Transformers for model inference
• Removes VLLM-based implementation and dependencies
• Updates router function to use HFChat as default fallback instead of returning None

Reviewed Changes

Copilot reviewed 7 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/tfbench/lm/utils.py	Updates router to import HFChat and use it as default fallback
src/tfbench/lm/_vllm.py	Removes entire VLLM implementation file
src/tfbench/lm/_hf.py	Adds new Hugging Face Transformers implementation
src/tfbench/experiment.py	Removes assertion check since router no longer returns None
src/tfbench/env.py	Simplifies environment handling using load_dotenv
src/main.py	Minor code reorganization moving result_dir creation
pyproject.toml	Replaces vllm dependency with transformers[torch]

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}