Feature Request: Add `ColabAILLMClient` for native Google Colab support

[cmccomb]

Description

Currently, users working within Google Colab environments often perform rapid prototyping and testing. To improve the developer experience and reduce boilerplate code for these users, I propose adding a formal ColabAILLMClient to the design-research-agents library.

This client would act as a bridge to the google.colab.ai service, allowing users to leverage the built-in Colab AI as an LLM provider without needing to manually define a custom class in every notebook.

Proposed Implementation

The implementation should align with existing clients in the llm_clients module.

1. Contract Compliance: The new class must inherit from (or adhere to) the LLMClient abstract base class/protocol to ensure compatibility with generate() and existing agent workflows.
2. Introspection: Implement the standard introspection surface:
   • default_model()
   • capabilities()
   • config_snapshot()
   • server_snapshot()
   • describe()
3. Graceful Dependency Management: Since google.colab is not available in standard local Python environments, the import should be handled safely (e.g., inside the class methods or protected by a try-except block). It would be ideal to include this under an "extra" dependency for packaging.

Example Usage

Once integrated, the workflow for a Colab user would be simplified to:

from design_research_agents.clients import ColabAILLMClient

client = ColabAILLMClient()
# Standard agent interaction follows...

Benefits

• Lower Barrier to Entry: Enables "zero-config" usage for students and researchers using Colab.
• Code Cleanup: Removes the need for users to maintain custom client logic in their own notebook cells.
• Standardization: Brings Colab-specific usage in line with other supported services like GeminiServiceLLMClient or OpenAIServiceLLMClient.