Add GPU utilization A/B test and fix cache-aware proxy bugs

- GPU monitor: 5s interval nvidia-smi sampling during benchmarks
- A/B test script: clean restart + monitor + benchmark for Combined vs PD-Sep
- Fixed proxy: await bootstrap init (race condition), normalized LB scoring
- Fixed port conflicts: proxy 9090 to avoid bootstrap 9000 clash

Key finding: PD-Sep GPU utilization is 40% of Combined (12.4% vs 30.5%)
- Decode GPUs: mean=7.8%, max=47% (memory-bound, compute wasted)
- Prefill GPUs: active only 17% of samples (bursty, idle between requests)
- Combined: 8 GPUs flexibly used, mean=30.5%, active=64%

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

scripts/cache_aware_proxy.py

@@ -132,7 +132,7 @@ async def lifespan(app: FastAPI):
             prefill_instances.append(InstanceState(url, bp))
         for url in global_args.decode:
             decode_instances.append(InstanceState(url))
-        asyncio.create_task(init_prefill_bootstrap(prefill_instances, app.state.ready))
+        await init_prefill_bootstrap(prefill_instances, app.state.ready)
         print(f"PD-Sep mode: {len(prefill_instances)}P + {len(decode_instances)}D")
 
     yield