[refactor] Refactoring AscendFusedMoE (#1229)

<!--  Thanks for sending a pull request!

BEFORE SUBMITTING, PLEASE READ
https://docs.vllm.ai/en/latest/contributing/overview.html

-->
### What this PR does / why we need it?
This PR is used for resolved [issue
1147](https://github.com/vllm-project/vllm-ascend/issues/1147)
1. Move fused_moe code into one file `fused_moe.py`.
2. Integrate branch conditions into function `get_fused_moe_state`.
<!--
- Please clarify what changes you are proposing. The purpose of this
section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster
reviews in your PR.

- Please clarify why the changes are needed. For instance, the use case
and bug description.

- Fixes #
-->

### Does this PR introduce _any_ user-facing change?
1. This PR has removed the env `VLLM_ENABLE_MC2`, because I think this
env is useless, we can make judgments based on the current scenario
without this env, it will only increase complexity.
2. This PR has removed the env `USING_LCCL_COM`, because this env has
already expired.
3. `additional_config.expert_tensor_parallel_size` has already expired,
and now we also use parameter `enable_expert_parallel`, consistent with
the vLLM.
<!--
Note that it means *any* user-facing change including all aspects such
as API, interface or other behavior changes.
Documentation-only updates are not considered user-facing changes.
-->

### How was this patch tested?
<!--
CI passed with new added/existing test.
If it was tested in a way different from regular unit tests, please
clarify how you tested step by step, ideally copy and paste-able, so
that other reviewers can test and check, and descendants can verify in
the future.
If tests were not added, please describe why they were not added and/or
why it was difficult to add.
-->

Signed-off-by: zzzzwwjj <1183291235@qq.com>

vllm_ascend/worker/model_runner_v1.py

@@ -348,15 +348,10 @@ class NPUModelRunner(LoRAModelRunnerMixin):
             self.init_torchair_graph_batch_sizes()
 
         if len(self.torchair_graph_batch_sizes) == 0:
-            #If MC2 is enabled, torchair_graph_batch_size should pad to tp_size
-            if envs_ascend.VLLM_ENABLE_MC2:
-                self.torchair_graph_batch_sizes = [
-                    self.scheduler_config.max_num_seqs
-                ]
-            else:
-                self.torchair_graph_batch_sizes = [
-                    1, self.scheduler_config.max_num_seqs
-                ]
+            # TODO(zzzzwwjj): check torchair_graph_batch_sizes init code
+            self.torchair_graph_batch_sizes = [
+                self.scheduler_config.max_num_seqs
+            ]
 
         torch._dynamo.cache_size.config.cache_size_limit += len(
             self.torchair_graph_batch_sizes)
@@ -569,10 +564,12 @@ class NPUModelRunner(LoRAModelRunnerMixin):
             self.input_batch.refresh_sampling_metadata()
 
     def _get_forward_metadata_across_dp(
-            self, batch_size: int, with_prefill: bool) -> tuple[int, bool]:
-        forward_metadata = torch.tensor([batch_size, with_prefill],
-                                        device="cpu",
-                                        dtype=torch.int32)
+            self, total_num_scheduled_tokens: int,
+            with_prefill: bool) -> tuple[int, bool]:
+        forward_metadata = torch.tensor(
+            [total_num_scheduled_tokens, with_prefill],
+            device="cpu",
+            dtype=torch.int32)
         dist.all_reduce(forward_metadata,
                         op=ReduceOp.MAX,
                         group=get_dp_group().cpu_group)
@@ -901,11 +898,11 @@ class NPUModelRunner(LoRAModelRunnerMixin):
         if self.dp_size > 1:
             max_num_tokens, with_prefill = self._get_forward_metadata_across_dp(
                 total_num_scheduled_tokens, with_prefill)
+            extra_builder_kwargs['max_num_tokens_across_dp'] = max_num_tokens
             extra_builder_kwargs['with_prefill_across_dp'] = with_prefill
 
         # Add graph_pad_size here
-        if envs_ascend.VLLM_ENABLE_MC2 or (self.torchair_graph_enabled
-                                           and not with_prefill):
+        if self.torchair_graph_enabled and not with_prefill:
             if self.dp_size > 1:
                 padded_batch_size = self.select_torchair_padded_batch_size(
                     max_num_tokens)
@@ -984,8 +981,7 @@ class NPUModelRunner(LoRAModelRunnerMixin):
         else:
             positions = self.positions[:num_input_tokens]
 
-        if (envs_ascend.VLLM_ENABLE_MC2
-                or self.torchair_graph_enabled) and not with_prefill:
+        if self.torchair_graph_enabled and not with_prefill:
             input_ids = self.input_ids[:padded_batch_size]
             positions = self.positions[:padded_batch_size]
 
@@ -1885,20 +1881,15 @@ class NPUModelRunner(LoRAModelRunnerMixin):
         return spec_token_ids
 
     def init_torchair_graph_batch_sizes(self):
+        start_graph_batch_size = 4
         tp_size = get_tensor_model_parallel_world_size()
-        batch_size_step = 8
-        largest_batch_size = 1
 
-        if envs_ascend.VLLM_ENABLE_MC2:
-            batch_size_step = max(batch_size_step, tp_size)
-            largest_batch_size = batch_size_step
-        while (largest_batch_size < 8):
-            self.torchair_graph_batch_sizes.append(largest_batch_size)
-            largest_batch_size *= 2
+        # NOTE: When use all2all | mc2, We need to slice the `num_tokens` dimension into `tp_size` blocks
+        start_graph_batch_size = max(start_graph_batch_size, tp_size)
 
-        while (largest_batch_size <= self.scheduler_config.max_num_seqs):
-            self.torchair_graph_batch_sizes.append(largest_batch_size)
-            largest_batch_size += batch_size_step
+        while (start_graph_batch_size <= self.scheduler_config.max_num_seqs):
+            self.torchair_graph_batch_sizes.append(start_graph_batch_size)
+            start_graph_batch_size *= 2
 
     def select_torchair_padded_batch_size(self, batch_size: int):
         selected_batch_size = self.max_num_reqs