Add Ollama integration with local LLMs and fix conversation history

[NeurArk]

Summary

• Add Ollama integration to support local language models
• Implement session management for persistent conversations
• Fix conversation history persistence using prev_result.to_input_list()
• Simplify chat history management with a single source of truth
• Add proper coroutine cleanup to avoid event loop warnings
• Add comprehensive test coverage for session management and Ollama integration

Features

• Support for local LLMs through Ollama integration
• UI toggle to switch between OpenAI and local models
• Fallback mechanisms when either service is unavailable
• Persistent conversation history across multiple exchanges

Test Plan

• Verify that conversation history is maintained between calls
• Confirm Ollama integration works with local models
• Run pytest suite to confirm all tests pass
• Ensure PDF reports are created without unnecessary clarification requests