[Edge][WP2] Add initialization timeline and cold-start profiler

* vllm_omni/engine/init_timeline.py: InitTimeline phase collector; engine
  phases (stages_init, g3_device_lock, g2_spawn/ready) and worker phases
  (proc_entry, engine_core_init, init_device, load_model/load_weights,
  talker_mtp_init, attn_prewarm, cudagraph_capture, talker_mtp_capture,
  code2wav_cudagraph, predictor_setup_compile/warmup/cudagraph) recorded via
  worker_phase/timed_phase; workers append JSONL under
  VLLM_OMNI_INIT_TIMELINE_DIR (inherited at spawn) and the engine merges them.
  Enabled by VLLM_OMNI_INIT_TIMELINE=<json> or log_stats; table logged at INFO,
  partial table on orchestrator startup timeout, suggested timeouts logged.
* omni_base.py: named deploy profiles default to stage_init_timeout=900 /
  init_timeout=1500 unless given explicitly.
* gpu_model_runner._init_talker_mtp tolerates max_cudagraph_capture_size=None;
  gpu_ar_model_runner.shutdown skips CUDA graph teardown on CPU.
* benchmarks/tts/init_profile.py: runs stream_latency_bench --init-only per
  configuration (default, cold_cache, eager_stage1, eager_both,
  parallel_stage_init, edge, ...) and tabulates phase durations.
* stream_latency_bench: --parallel-stage-init / --deploy-profile flags; the
  resource poller no longer shadows Thread._stop and skips GPU sampling on CPU.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

vllm_omni/worker/gpu_ar_model_runner.py

@@ -45,6 +45,7 @@ from vllm.v1.worker.utils import is_residual_scattered_for_sp
 from vllm_omni.data_entry_keys import flatten_payload
 from vllm_omni.distributed.omni_connectors.kv_transfer_manager import OmniKVTransferManager
 from vllm_omni.distributed.omni_connectors.utils.config import stage_sends_async_output
+from vllm_omni.engine.init_timeline import worker_phase
 from vllm_omni.model_executor.duplex_sampling import DuplexSamplingRunnerMixin
 from vllm_omni.outputs import OmniModelRunnerOutput
 from vllm_omni.utils.mm_outputs import (
@@ -469,8 +470,10 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin, Duplex
         return engine_output_type, downstream_req_ids
 
     def capture_model(self) -> int:
-        result = super().capture_model()
-        self._capture_talker_mtp_graphs()
+        with worker_phase("cudagraph_capture"):
+            result = super().capture_model()
+        with worker_phase("talker_mtp_capture"):
+            self._capture_talker_mtp_graphs()
         return result
 
     def shutdown(self) -> None:
@@ -522,11 +525,13 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin, Duplex
         # 5. Release all CUDA graphs unconditionally (upstream only does this
         #    on ROCm; on CUDA the graphs are only freed by Python GC during
         #    interpreter shutdown, which is too late to prevent memory spikes).
-        from vllm.compilation.breakable_cudagraph import BreakableCUDAGraphWrapper
-        from vllm.compilation.cuda_graph import CUDAGraphWrapper
+        #    Skipped on CPU, where no graph wrapper is ever instantiated.
+        if getattr(self.device, "type", "cuda") != "cpu":
+            from vllm.compilation.breakable_cudagraph import BreakableCUDAGraphWrapper
+            from vllm.compilation.cuda_graph import CUDAGraphWrapper
 
-        CUDAGraphWrapper.clear_all_graphs()
-        BreakableCUDAGraphWrapper.clear_all_graphs()
+            CUDAGraphWrapper.clear_all_graphs()
+            BreakableCUDAGraphWrapper.clear_all_graphs()
 
         # 6. Delegate to upstream shutdown (model = None, KV caches, workspace).
         super().shutdown()