[Minor] Zero-initialize attn output buffer (#19784)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

Author: WoosukKwon

vllm/attention/layer.py

@@ -209,7 +209,7 @@ def forward(
         if self.use_output:
             output_shape = (output_shape
                             if output_shape is not None else query.shape)
-            output = torch.empty(output_shape,
+            output = torch.zeros(output_shape,
                                  dtype=query.dtype,
                                  device=query.device)
             hidden_size = output_shape[-1]