Padding in permutation for GroupedTensor operations

[ptrendx]

Enable explicit support and optimization for padding within permutation operations on GroupedTensors—aimed at improving efficiency in batch processing and layout transformations.

[tdophung]

Done with #1921 merged