Text this: VP-SFDA: Visual Prompt Source-Free Domain Adaptation for Cross-Modal Medical Image