Text this: Early‐Life Fuel Cell Degradation Prediction via Differential‐Integral Enhanced Hybrid Networks With Cross‐Stack‐Operation Transfer Learning