[Refactor] Fix AttentionMaskBuilder singleton and remove redundant pcp_prefill_mask (#4870)

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

This PR fixes the `AttentionMaskBuilder` singleton initialization issue
introduced in PR #4779 and removes the unused `pcp_prefill_mask` field.

### Background

After PR #4779 made `AttentionMaskBuilder` a singleton with `@singleton`
decorator, the class constructor now requires a `device` parameter.
However, two initialization sites were still using the old parameterless
constructor, causing failures.

### Changes

1. **Fix singleton initialization**
- Fixed `AttentionMaskBuilder()` → `AttentionMaskBuilder(self.device)`
in `AscendMLAMetadataBuilder.__init__()`
- Fixed `AttentionMaskBuilder()` → `AttentionMaskBuilder(self.device)`
in `AscendAttentionMetadataBuilder.__init__()`

2. **Remove unused field**
- Removed `pcp_prefill_mask` field from
`AscendPrefillContextParallelMetadata` (never used in codebase)
   - Updated related test assertions

### Related

- Issue #5463
- PR #4779 (Unify all mask generation methods)
- PR #5389 (Make AttentionMaskBuilder singleton)

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

No. This is an internal refactoring.

## How was this patch tested?

- ✅ Local testing: No linter errors
- ✅ Unit tests for attention modules verified
- ⏳ CI pipeline

Signed-off-by: lico67373 <918688502@qq.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>

vllm_ascend/attention/attention_mask.py

@@ -13,6 +13,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import torch
+from vllm.distributed import get_pcp_group
+
+from vllm_ascend.platform import ModelConfig
+from vllm_ascend.utils import singleton
 
 
 def _generate_attn_mask(max_seq_len, dtype):
@@ -29,6 +33,7 @@ def _generate_attn_mask(max_seq_len, dtype):
     return attn_mask
 
 
+@singleton
 class AttentionMaskBuilder:
 
     def __init__(self, device: torch.device):
@@ -82,4 +87,16 @@ class AttentionMaskBuilder:
                 triu_mask = torch.triu(mask, diagonal=1).to(self.device)
                 tril_mask = torch.tril(mask, -sliding_window).to(self.device)
                 self.swa_mask = triu_mask + tril_mask
-        return self.swa_mask
+        return self.swa_mask
+
+    def get_attention_mask(self, model_config: ModelConfig):
+        if model_config.runner_type == "pooling":
+            return self.get_attn_mask(2048, torch.bool)
+
+        return self.get_splitfuse_attn_mask()
+
+    def get_final_mla_mask(self, model_config: ModelConfig):
+        if get_pcp_group().world_size > 1:
+            return self.get_pcp_mla_mask(model_config.dtype)
+        # Prefill stages use 512x512 mask with appropriate dtype
+        return self.get_mla_mask(model_config.dtype)