[Feat][Graph] Support MTP for ACL Graph (#2932)

### What this PR does / why we need it?
This PR depends on the merge of #2707 and has adapted the aclgraph
functionality to support MTP.

### How was this patch tested?


- vLLM version: v0.10.2
- vLLM main:
2b85697031

---------

Signed-off-by: xuyexiong <xuyexiong@huawei.com>

vllm_ascend/attention/mla_v1.py

@@ -187,7 +187,14 @@ class AscendMLAMetadataBuilder:
                            self.block_size - 1) // self.block_size
         self.chunked_prefill_enabled = scheduler_config.chunked_prefill_enabled
 
+        self.speculative_config = vllm_config.speculative_config
         self.decode_threshold = 1
+        if self.speculative_config:
+            spec_token_num = self.speculative_config.num_speculative_tokens
+            self.decode_threshold += spec_token_num
+            assert self.decode_threshold <= 16, f"decode_threshold exceeded \
+                npu_fused_infer_attention_score TND layout's limit of 16, \
+                got {self.decode_threshold}"
 
         if self.chunked_prefill_enabled:
             self.chunked_prefill_workspace_size = min(
@@ -275,7 +282,6 @@ class AscendMLAMetadataBuilder:
         num_actual_tokens = common_attn_metadata.num_actual_tokens
         query_start_loc = common_attn_metadata.query_start_loc
         query_start_loc_cpu = common_attn_metadata.query_start_loc_cpu
-        # TODO(xyx): remove the if condition after mla supports torch mode speculative decoding
         num_decodes, num_prefills, num_decode_tokens, num_prefill_tokens = \
             split_decodes_and_prefills(common_attn_metadata, decode_threshold=self.decode_threshold)
         assert num_decodes + num_prefills == num_reqs