Light Up the Shadows: Enhance Long-Tailed Entity Grounding with Concept-Guided Vision-Language Models