Reinforcement Learning in Linear Quadratic Deep Structured Teams with Performance Guarantee