[MoE Refactor] Introduce MoERunner abstraction and move execution logic from FusedMoE to DefaultMoERunner (#32344)

Signed-off-by: Bill Nell <bnell@redhat.com>

vllm/model_executor/layers/quantization/mxfp4.py

@@ -900,6 +900,7 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
         x: torch.Tensor,
         topk_weights: torch.Tensor,
         topk_ids: torch.Tensor,
+        shared_experts_input: torch.Tensor | None,
     ) -> torch.Tensor | tuple[torch.Tensor, torch.Tensor]:
         assert not self.is_monolithic
         if layer.enable_eplb: