[v0.18.0][Bugfix] fix ds3.2 dcp mtp (#7681)

### What this PR does / why we need it?
Fixed the issue where the DCP overlaps the MTP scenario in the ds3.2
scenario.

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

### How was this patch tested?

cherry-pick from: https://github.com/vllm-project/vllm-ascend/pull/7617

Signed-off-by: weiguihua2 <weiguihua2@huawei.com>

vllm_ascend/attention/mla_v1.py

@@ -174,6 +174,7 @@ class AscendMLAMetadata:
     slot_mapping: torch.Tensor
     query_start_loc: torch.Tensor
     seq_lens: torch.Tensor
+    seq_lens_cpu: torch.Tensor
     block_tables: torch.Tensor
 
     # New for MLA (compared to FlashAttention)
@@ -457,6 +458,7 @@ class AscendMLAMetadataBuilder(MLACommonMetadataBuilder[AscendMLAMetadata]):
             query_start_loc=query_start_loc,
             block_tables=self.block_table,
             seq_lens=self.seq_lens,
+            seq_lens_cpu=self.seq_lens,
         )
 
     def build_chunked_metadata(