Prompts & Provenance: Re-situating Place in Text-to-Image Models