Characterizing large language model generative artificial intelligence variability in the production of objective structured clinical examination stations (Preprint)