[main][refactor] Refactoring forward_context and model_runner_v1 (#1979)

### What this PR does / why we need it?

A refactoring of forward_context and model_runner_v1, add some context
which is necessary in model inference into forward_context, and refactor
dummy_run logic, make it more reasonable.
Some details for this PR:

Add `ascend_forward_context`;
Update mc2_v2 op, and support `active_mask` param;
Update scripts in examples dir;
refactor `dummy_run` logic;
Add soc_version for A2 and A3;

### Does this PR introduce _any_ user-facing change?

No change at user-facing.

### How was this patch tested?


- vLLM version: v0.10.0
- vLLM main:
57c22e57f9

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

vllm_ascend/attention/mla_v1.py

@@ -126,9 +126,6 @@ class AscendMLAMetadata:
     # For logging.
     num_input_tokens: int = 0  # Number of tokens including padding.
 
-    max_num_tokens_across_dp: int = 0
-    with_prefill_across_dp: bool = False
-
     query_lens: Optional[list[int]] = None
     # The dimension of the attention heads
     head_dim: Optional[int] = None
@@ -302,8 +299,8 @@ class AscendMLAMetadataBuilder:
 
         return graph_block_tables[:num_seqs, :max_blocks]
 
-    def build_dummy(self, num_reqs: int,
-                    num_actual_tokens: int) -> AscendMLAMetadata:
+    def build_torchair_graph_dummy(
+            self, num_reqs: int, num_actual_tokens: int) -> AscendMLAMetadata:
         device = self.runner.device
         _, max_blocks = self.runner.graph_block_tables.shape
         block_table = torch.zeros((num_reqs, max_blocks),
@@ -353,8 +350,6 @@ class AscendMLAMetadataBuilder:
         num_actual_tokens: int,
         max_query_len: int,
         graph_pad_size: int = -1,
-        max_num_tokens_across_dp: int = 0,
-        with_prefill_across_dp: bool = False,
         query_start_loc: torch.Tensor = None,
     ) -> AscendMLAMetadata:
         assert self._num_decodes + self._num_prefills == num_reqs
@@ -498,8 +493,6 @@ class AscendMLAMetadataBuilder:
             query_start_loc=query_start_loc,
             block_tables=block_table,
             seq_lens=seq_lens,
-            max_num_tokens_across_dp=max_num_tokens_across_dp,
-            with_prefill_across_dp=with_prefill_across_dp,
         )