Add OpenRouter API Support

[morning3tar]

This pull request adds support for using OpenRouter.ai models as LLM backends in AgentLaboratory.
It allows users to select any OpenRouter-supported model (including free models) via the YAML config and routes requests to the OpenRouter API.
OpenRouter API logic is separated into its module (openrouter_inference.py) for maintainability.
For unknown or OpenRouter models, the code defaults to the cl100k_base tokenizer to avoid errors.

How to Use

• Add your OpenRouter API key to your YAML config
• Specify the backend model in your YAML config: llm-backend: "openrouter/deepseek/deepseek-chat-v3-0324:free"

Run AgentLaboratory as usual.