Tell2Reg: Establishing spatial correspondence between images by the same language prompts