【Feature】refactor npu_modelrunner for profile_run (#4993)

### What this PR does / why we need it?
(1)refactor npu_model_runner for profile_run
(2) move _select_moe_comm_method to ascend_forward_context
(3) delete _init_model_kwargs in npu_model_runner

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
Na
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: zhenwenqi2024 <zhenwenqi_2022@qq.com>
Signed-off-by: zhenwenqi2024 <155598497+zhenwenqi2024@users.noreply.github.com>

vllm_ascend/spec_decode/eagle_proposer.py

@@ -123,10 +123,9 @@ class EagleProposer(Proposer):
                   aclgraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
                   batch_descriptor=None,
                   dummy_compute_logits=lambda hidden_states: None):
-        moe_comm_type = self.runner._select_moe_comm_method(num_tokens)
         with set_ascend_forward_context(None,
                                         self.vllm_config,
-                                        moe_comm_type=moe_comm_type,
+                                        in_profile_run=True,
                                         num_tokens=num_tokens):
             self.model(
                 input_ids=self.input_ids[:num_tokens],
@@ -458,15 +457,12 @@ class EagleProposer(Proposer):
         else:
             num_input_tokens = num_tokens
 
-        moe_comm_type = self.runner._select_moe_comm_method(num_input_tokens)
-
         # copy inputs to buffer for cudagraph
         self.positions[:num_tokens] = target_positions.to(device)
         self.hidden_states[:num_tokens] = target_hidden_states
         attn_metadata.block_tables = block_table.to(device)
         with set_ascend_forward_context(attn_metadata,
                                         self.vllm_config,
-                                        moe_comm_type=moe_comm_type,
                                         num_tokens=num_input_tokens):
             last_hidden_states, hidden_states = self.model(
                 input_ids=self.input_ids[:num_input_tokens],
@@ -498,8 +494,6 @@ class EagleProposer(Proposer):
         else:
             input_batch_size = batch_size
 
-        moe_comm_type = self.runner._select_moe_comm_method(input_batch_size)
-
         attn_metadata.num_actual_tokens = batch_size
         attn_metadata.max_query_len = 1
         attn_metadata.query_start_loc = self.arange[:batch_size + 1]
@@ -575,7 +569,6 @@ class EagleProposer(Proposer):
             # Run the model.
             with set_ascend_forward_context(attn_metadata,
                                             self.vllm_config,
-                                            moe_comm_type=moe_comm_type,
                                             num_tokens=input_batch_size):
 
                 last_hidden_states, hidden_states = self.model(