feat: Add compatibility for Openrouter Endpoints

[akshaypardhanani]

Overview

Changes here:

1. Add OpenRouter API support alongside existing OpenAI compatibility.
2. Add uv as a package and environment manager

Key Features

1. OpenRouter API Integration
   Open router can be called through the OpenAI sdk. We set the base url to the openrouter endpoint to enable this.
   This enables multiple models that can be used. Through the development of the changes I have tested with the following:

• Qwen3 coder: Openrouter model is qwen/qwen3-coder:free
• Kimi dev: Openrouter model is moonshotai/kimi-dev-72b:free
• GLM z1 32B: Openrouter model is thudm/glm-z1-32b:free 👈 this is not available for free anymore

2. API Key Management
   Created amadeusgpt/utils/api_key_util.py for centralized API key handling
   Supports OPENAI_API_KEY, and OPENROUTER_API_KEY, to be specified as environment variables

Testing

• Tested out the package through the notebooks. I have added version suffixed with _openrouter to replicate the existing demos with an openrouter API KEY
• Tested manually via the streamlit app - This was breaking due to a compatibility mismatch between the latest version(s) of streamlit and streamlit-drawable-canvas (which is now archived and read only). As a result I minimally cloned it locally to be used as a local module.
• Ran unit tests in the tests folder - Tests with SuperAnimal in the test_superanimal.py suite are not fully tested yet since they use a vision language model (to my understanding) I am working on the fixes needed to get that working (which is why I am marking this Pull request as Draft for now).