Similar Items: Multi‐Grained Vision–Language Alignment for Domain Generalised Person Re‐Identification