Web crawler that converts documentation sites to clean Markdown and JSONL for LLM ingestion (RAG, training data).