Towards a Foundation Model for Physics-Informed Neural Networks: Multi-PDE Learning with Active Sampling