Breaking the Curse of Dimensionality in Multiagent State Space: A Unified Agent Permutation Framework