Add Ollama integration for local LLM Inference

[AnandMayank]

🚀 Add Ollama Integration for Local LLM Inference

This PR introduces local Ollama support that achieves better GPT scores and GPT_std metrics in zero-shot evaluation while enabling privacy-focused, cost-free operation.

✨ Key Improvements

• 🏆 Better Performance: Improved GPT scores and GPT_std without fine-tuning
• 🚀 Immediate Start: No Hugging Face tokens or authentication required
• 🔧 Fixes Finetune Issues: Resolves existing finetune command errors (addresses An error occurred while running the Python explainer/main.exe -- mode finetune -- dataset Yelp command #12, Error when running finetune cmd #13)
• 🔒 Privacy-First: Enable completely local operation with no external API calls
• 💰 Cost-Free: Eliminate API fees for researchers and developers
• ✅ Full Compatibility: Maintain all existing evaluation metrics and functionality
• 🛠️ Easy Setup: Include automated setup scripts and comprehensive documentation

🎯 Setup Advantages

• No Account Required: Skip Hugging Face registration and token setup
• No Authentication: Eliminate common API key configuration issues
• Instant Access: Download, install, and run immediately
• Offline Capable: Works without internet connection after initial model download

🐛 Issues Resolved

• Fixes finetune command errors that users have been experiencing
• Eliminates Hugging Face authentication and token-related issues
• Provides stable local alternative when cloud APIs are unavailable

📈 Performance Benefits

• Zero-Shot Excellence: Better evaluation metrics without requiring fine-tuning
• No Training Needed: Superior results achieved immediately with local models
• Reliable Operation: Local inference eliminates API timeout and rate limit issues