⚡ perf: pre-allocate GQA buffers in XAttention policy

Add alloc_policy_metadata() method to SparsePolicy base class for
pre-allocating GPU buffers during initialization. This avoids
dynamic memory allocation during forward pass.

Changes:
- Add alloc_policy_metadata() to SparsePolicy base class
- Implement GQA buffer pre-allocation in XAttentionBSAPolicy
- Call alloc_policy_metadata() in model_runner for GPU-only mode
- Modify compute_prefill() to reuse pre-allocated buffers
- Add --gpu-util parameter to bench.py

Memory savings:
- Previously: 2x GQA expansion (~2GB for 64K)
- Now: 1x pre-allocated buffer (~1GB for 64K, reused)

Tested:
- GPU-only 32K: 5602 tok/s (512MB pre-allocated)
- GPU-only 64K: 4821 tok/s (1GB pre-allocated, gpu_util=0.7)
- Offload Full: PASSED (no changes to offload path)
- Offload XAttention: PASSED (uses compute_chunked_prefill)

Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)

Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>

bench.py

@@ -56,6 +56,8 @@ def main():
                         help="Sparse policy: full (FullAttention), xattn (XAttention+BSA)")
     parser.add_argument("--enable-policy", action="store_true",
                         help="Enable sparse policy routing (FullAttentionPolicy by default)")
+    parser.add_argument("--gpu-util", type=float, default=0.9,
+                        help="GPU memory utilization (default: 0.9)")
     args = parser.parse_args()
 
     path = os.path.expanduser(args.model)
@@ -78,6 +80,7 @@ def main():
         max_model_len=max_len,
         max_num_batched_tokens=max_len,
         sparse_policy=sparse_policy,
+        gpu_memory_utilization=args.gpu_util,
     )
 
     # Warmup