RDLNN-DCRAS: Non-Destructive Continual Learning via Hebbian Lateral Voting in a Biologically-Grounded Liquid Neural Cognitive Architecture | Synapse