Safe Distributed Learning-Enhanced Predictive Control for Multiple Quadrupedal Robots