Text this: Performance of vision language models for optic disc swelling identification on fundus photographs