Antisymmetric polyspectral indices for high-order neural interactions: general theory and a fourth-order proof of concept