[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/attention_v1.py

@@ -168,6 +168,7 @@ class AscendMetadata:
     # should simplified these parameters once attention schema in vLLM-Ascend
     # is unified.
     seq_lens: torch.Tensor = None
+    seq_lens_cpu: torch.Tensor = None
     seq_lens_list: list[int] = None  # type: ignore
     actual_seq_lengths_q: list[int] = None  # type: ignore
 
@@ -307,6 +308,7 @@ class AscendAttentionMetadataBuilder(AttentionMetadataBuilder[AscendMetadata]):
             block_tables=block_table,
             query_start_loc=query_start_loc,
             seq_lens=seq_lens,
+            seq_lens_cpu=seq_lens,
             seq_lens_list=seq_lens.tolist(),
             max_query_len=common_attn_metadata.max_query_len,
             actual_seq_lengths_q=query_start_loc_cpu[1:].tolist(),