Add Ollama local VLM testing guide

[kiyeonjeon21]

What

Document how to use Ollama with parsemux for free, local VLM image description.

Content needed

1. Install Ollama + pull qwen2.5vl:7b
2. Run: `parsemux parse doc.pdf --extract-images --describe-images` (auto-detects Ollama when no API key)
3. Performance expectations (speed, quality vs cloud)
4. Add to README or create docs/ollama-guide.md

Context

Ollama is the only zero-cost VLM option. Great for privacy-conscious users.