A Multimodal PDE Foundation Model for Prediction and Scientific Text Descriptions