Logically Consistent Language Models via Neuro-Symbolic Integration