Add structured case JSON (cases/): grounded caption facts, dispositions, citations

Add positioned text blocks (layout/) with four coordinate spaces

Add OCR'd state court opinions (Unlimited-OCR via vLLM)