Does Vision-and-Language Pretraining Improve Lexical Grounding?