CAL-L4: The Meta-Inference Layer — L4 Efficiency Hypothesis and the Representational Convergence Conjecture | Synapse