A Multi-dimensional Deep Structured State Space Approach to Speech Enhancement Using Small-footprint Models