IVLMap: Instance-Aware Visual Language Grounding for Consumer Robot Navigation