Catalogue Grounded Multimodal Attribution for Museum Video