WMNav: Integrating Vision-Language Models into World Models for Object Goal Navigation