Unify Variables in Neural Scaling Laws for General Audio Representations via Embedding Effective Rank