Scene-centric vs. Object-centric Image-Text Cross-modal Retrieval: A Reproducibility Study