Expand Ollama docs page (local models, model management, GPU config)

[AmitoVrito]

Expand docs/llms/ollama.md with:

• Installing and starting Ollama
• Pulling models (llama3, mistral, codellama, phi3)
• Custom model URLs for remote Ollama
• GPU memory requirements per model size
• Streaming with local models
• No-API-key setup for CI/testing
• Performance tips (keep_alive, num_ctx)

[AmitoVrito]

Completed in PR #99 — ollama.md expanded to 200 lines with GPU config, model management, platform install.