[main] [refactor] refactor common_fused_moe.py (vllm-project#2706)

### What this PR does / why we need it?
1. Move prepare/finalize operation from moe_comm_method to
/ops/moe/fused_moe_prepare_and_finalize
2. Adapt to token_dispatcher in moe_comm_method
3. Move
moe_comm_method/experts_selector/token_dispatcher/fused_moe_prepare_and_finalize
to /ops/moe
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
e2e & ut

- vLLM version: v0.10.1.1
- vLLM main:
vllm-project/vllm@f4962a6

Signed-off-by: weichen <calvin_zhu0210@outlook.com>
Signed-off-by: Pr0Wh1teGivee <calvin_zhu0210@outlook.com>
Co-authored-by: weijinqian0 <12153182+weijinqian0@users.noreply.github.com>

Author: 2 people

tests/e2e/singlecard/ops/test_fused_moe.py

@@ -28,9 +28,8 @@
 import torch_npu
 from vllm.model_executor.layers.activation import SiluAndMul
 
-from vllm_ascend.ops.layers.experts_selector import select_experts
-from vllm_ascend.ops.moe_dispatcher.token_dispatcher import \
-    TokenDispatcherWithAllGather
+from vllm_ascend.ops.moe.experts_selector import select_experts
+from vllm_ascend.ops.moe.token_dispatcher import TokenDispatcherWithAllGather
 
 NUM_EXPERTS = [8, 64]
 EP_SIZE = [1]
@@ -209,7 +208,7 @@ def test_select_experts(
                                  dtype=torch.int32)
         custom_routing_function.return_value = (mock_weights, mock_ids)
 
-    with patch("vllm_ascend.ops.layers.experts_selector._native_grouped_topk"
+    with patch("vllm_ascend.ops.moe.experts_selector._native_grouped_topk"
                ) as mock_native_grouped_topk:
         mock_native_grouped_topk.side_effect = lambda x, num_groups, k: torch.randn_like(
             x)