A Unifying Perspective on Multicalibration: Game Dynamics for Multi-Objective Learning