Towards Unified Interactive Visual Grounding in The Wild