[PyTorch] Add ops for MoE grouped MLP (#2664)

* Add ops for MoE grouped MLP

Signed-off-by: Tim Moon <tmoon@nvidia.com>

* Move testing utility functions to util submodule

Signed-off-by: Tim Moon <tmoon@nvidia.com>

* Tweak docs

Signed-off-by: Tim Moon <tmoon@nvidia.com>

* Change order of tensor compatibility checks in noop_cat

Review suggestion from @ptrendx.

Signed-off-by: Tim Moon <tmoon@nvidia.com>

* Add support for GLU interleaving in clamped SwiGLU

Signed-off-by: Tim Moon <tmoon@nvidia.com>

---------

Signed-off-by: Tim Moon <tmoon@nvidia.com>

Author: timmoon10