VAuLT: Augmenting the Vision-and-Language Transformer for Sentiment Classification on Social Media