Text this: Cross‐Method Explanation Stability Under Prediction‐Preserving Perturbations in Explainable AI