fix: accept cache_dtype_str in get_kv_cache_shape

vllm_npu/attention/attention_v1.py

@@ -83,6 +83,7 @@ class AscendAttentionBackend(AttentionBackend):
         block_size: int,
         num_kv_heads: int,
         head_size: int,
+        **kwargs,
     ) -> Tuple[int, int, int, int]:
         """KV cache shape: (num_blocks, block_size, num_kv_heads, head_size).
 
@@ -91,6 +92,7 @@ class AscendAttentionBackend(AttentionBackend):
         """
         return (num_blocks, block_size, num_kv_heads, head_size)
 
+
     @staticmethod
     def swap_blocks(
         src_kv_cache: List[torch.Tensor],