Add support for streaming llamaindex agents

[tuntisz]

Add LlamaIndex Agent Framework Support

This PR implements comprehensive support for LlamaIndex as an agent framework in the AI Data Stream project, enabling streaming AI responses that comply with Vercel's AI SDK Data Stream Protocol.

Key Changes

Core Implementation:

• Added LlamaIndexStreamer class that implements both sync and async streaming interfaces
• Added LlamaIndexMessageParser for converting between internal and LlamaIndex message formats
• Added LlamaIndexStreamConverter for transforming LlamaIndex responses to data stream protocol format
• Added DataStreamToolCallDelta class to handle incremental tool call updates

Features:

• Full support for streaming text responses
• Tool call and tool result handling
• Memory management and conversation history preservation
• Both synchronous and asynchronous interfaces
• Proper protocol ordering (start → content/tool calls → finish)

Testing:

• Comprehensive scenario-based test suite covering:
  • Simple text streaming
  • Tool calls (single and parallel)
  • Multi-turn conversations
  • Edge cases and error handling
• Mock utilities for testing without real LlamaIndex agents

Configuration:

• Added optional llama-index dependency group in pyproject.toml
• Updated .gitignore to exclude Claude Swarm configuration files
• Updated dependency versions (aiohappyeyeballs, aiohttp)

Documentation

• Added detailed README in ai_datastream/agent/llamaindex/
• Includes usage examples, architecture details, and integration guides

[ofek1weiss]

@tuntisz if you could add this framework as a bullet in the main README file that would be awesome

[ofek1weiss]

Just make sure to fix the pre-commit issues