fix: add out_buf to attn_paged_decode for CUDA graph capture compatibility

- Pre-allocate decode_out in InferenceWorkspace so attn_paged_decode does not call torch::empty inside graph capture
- Wire decode_out through KVCache, PagePool.bind_tasks, and CudaBackend.fwd_decode
- Run live forward before graph capture to get valid output (graph pool memory is zeroed after capture block exits)
- Greedy generation with graph replay is bit-exact across all batch sizes
- Decode speedups vs no-graph: B=1 2.09x, B=4 1.80x, B=8 1.94x, B=16 1.76x

astrai/inference/core/workspace.py

@@ -108,6 +108,14 @@ class InferenceWorkspace:
             device=device,
         )
 
+        # Decode output buffer (graph-safe pre-alloc).  Shape matches the
+        # decode kernel's output: [batch, q_head, head_dim].
+        self.decode_out = torch.empty(
+            (max_batch_size, max_q_heads, head_dim),
+            dtype=dtype,
+            device=device,
+        )
+
     def decode_buffers(self, batch: int, q_heads: int):
         """Return ``(o_part, ml_part)`` view sliced to live dimensions."""
         return (