Evaluation of Multilingual Ability to Use Spatial Deictic Expressions in Vision-Language Models