docs: add modal-gpu-ez integration

[algocean (algocean1204)]

Summary

Add documentation for langchain-modal-gpu-ez, a LangChain integration that wraps modal-gpu-ez for running HuggingFace models on Modal serverless GPUs.

Note: This is different from the existing langchain-modal (sandbox). langchain-modal-gpu-ez provides LLM and Embeddings interfaces for direct model inference.

What's added

• Provider page (providers/modal_gpu_ez.mdx): Overview, installation, setup
• LLM page (llms/modal_gpu_ez.mdx): ModalGpuEzLLM usage with GPU selection guide
• Embeddings page (text_embedding/modal_gpu_ez.mdx): ModalGpuEzEmbeddings usage with vector store example

Package features

• One-liner inference: llm.invoke("Hello") on T4/A100/H100 GPUs
• Auto GPU selection based on model size
• Local fallback (CUDA/MPS/CPU) at no cost
• Built-in cost tracking and budget management

Links

• PyPI: https://pypi.org/project/langchain-modal-gpu-ez/
• PyPI (core): https://pypi.org/project/modal-gpu-ez/

[algocean (algocean1204)]

Hi maintainers! 👋

This is my very first open-source PR, and as a university student, I'm super excited to contribute to the LangChain ecosystem.

Since I'm still learning, I would highly appreciate any direct, realistic, and strict feedback regarding the code structure, conventions, or the overall integration approach. Please don't hold back—I really want to learn and improve this to perfectly match LangChain's standards!

Thank you so much for your time reviewing this.