[BugFix] Xlite: Bypass the padding of the graph mode in non-MTP cases to obtain the correct decode num. (#5711)

### What this PR does / why we need it?
This PR fixes a bug in Xlite
backend(https://atomgit.com/openeuler/GVirt/issues/1), The direct cause
of the problem is that the XModel::PrepareAttn function obtained an
illegal number of tokens to be inferred, -540. This illegal value is due
to the padding feature of inference in graph mode and the residual state
across steps. This issue is triggered when a prefill request is newly
added in a step and a decode ends simultaneously. It is first fixed
using num_decode_tokens instead of attn_metadata.num_decodes.
1. In graph mode, vllm_ascend has padding characteristics. In the
_prepare_inputs function, if the number of tokens to be inferred is less
than the set threshold (8 in this case), the attn_metadata.num_decode
array will be expanded to 8.
2. Meanwhile, vllm_ascend uses the class variable self.query_start_loc
of NPUModelRunner to record the tokens to be inferred. Due to poor
coordination with the graph mode padding mechanism when crossing steps,
in some cases (such as when a decode request is completed in a certain
step and a new prefill request is added at the same time), negative
values may be calculated for attn_metadata.query_lens.
3. After type conversion, the negative values in query_lens cause an
overflow. Xlite detects that the number of tokens to be inferred for the
decode request is too large and triggers a "decode len too long" alert.

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

### How was this patch tested?
Same with https://atomgit.com/openeuler/GVirt/issues/1
- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef

Signed-off-by: wwwumr <1127858301@qq.com>

vllm_ascend/xlite/xlite.py

@@ -255,7 +255,14 @@ class XliteWrapper:
         ]
 
         if not with_prefill or self.full_mode:
-            batch = attn_metadata.num_prefills + attn_metadata.num_decodes
+            # TODO: When vllm_ascend enables graph mode, attn_metadata.num_decodes
+            # will be padded in decode requests. Therefore, it is first fixed using
+            # num_decode_tokens. However, in the future, when MTP is enabled, there
+            # may be cases where a single request involves multiple tokens, which
+            # will need to be solved.
+            num_decodes = attn_metadata.num_decode_tokens
+            num_prefills = attn_metadata.num_prefills
+            batch = num_prefills + num_decodes
             seq_lens = attn_metadata.seq_lens[:batch]
             seq_tensor = torch.cat([
                 torch.tensor([0]),
@@ -269,9 +276,9 @@ class XliteWrapper:
             xlite_attn_metadata = ModelAttnMeta()
             xlite_attn_metadata.lens = query_lens.tolist()
             xlite_attn_metadata.cached_lens = cached_lens.tolist()
-            xlite_attn_metadata.is_prefills = [
-                False
-            ] * attn_metadata.num_decodes + [True] * attn_metadata.num_prefills
+            xlite_attn_metadata.is_prefills = [False] * num_decodes + [
+                True
+            ] * num_prefills
             xlite_attn_metadata.block_tables = attn_metadata.block_tables.cpu(
             ).tolist()