[Perf][V1] Fully overlap model execution (#2783)

This PR is based on top of
[#23569](https://github.com/vllm-project/vllm/pull/23569) and
[#24219](https://github.com/vllm-project/vllm/pull/24219).

### What this PR does / why we need it?
This PR allows the model runner to function asynchronously when using
async scheduling. This allows full overlap of the cpu operations
(including prepare_inputs) and the model forward pass. This diff is
functional and does not support speculative decoding, PP, or guided
decoding.

Expected speedup is 5-10% over the current async scheduling.

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
server
```
python -m vllm.entrypoints.openai.api_server --model=Qwen3-32B\
	--trust-remote-code --enforce-eager \
	--distributed-executor-backend=mp \
	-tp=4 \
	--port 8006 \
	--max-model-len 32000 \
	--block-size 128 \
	--gpu-memory-utilization 0.99
```
client
```
python $TEST_PY --backend vllm --trust-remote-code --model Qwen3-32B \
  --dataset-name random --random-input-len 2048 --random-output-len 2048 \
  --ignore-eos\
  --num-prompts 48 --max-concurrency 48  --request-rate inf --temperature 0 \
  --metric-percentiles 90  --base-url http://localhost:8006 --save-result \
  --result-dir $PROFILER_DIR
```

benchmark test based on Qwen3-32B TPOT result:
||forward async| scheduler async |sync|
|-|-|-|-|
|avg|41.73|41.86|44.20|
|improve0|0.3%|0|0|
|improve1|5.58%|0|0|

benchmark test based on Qwen2___5-VL-7B-Instruct TPOT result:
||forward async|sync|
|-|-|-|
|avg|23.22|29.16|
|improve|20.3%|0|


- vLLM version: main
- vLLM main:
e93f4cc9e3

Signed-off-by: jiangpeng36 <jiangpeng36@huawei.com>
Signed-off-by: Ronald1995 <ronaldautomobile@163.com>
Co-authored-by: jiangpeng36 <jiangpeng36@huawei.com>
Co-authored-by: Ronald1995 <ronaldautomobile@163.com>

vllm_ascend/worker/worker_v1.py

@@ -18,7 +18,7 @@
 #
 
 import copy
-from typing import Optional
+from typing import Optional, Union
 
 import torch
 import torch.nn as nn
@@ -38,8 +38,8 @@ from vllm.tasks import SupportedTask
 from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE, GiB_bytes
 from vllm.v1.core.sched.output import SchedulerOutput
 from vllm.v1.kv_cache_interface import KVCacheConfig, KVCacheSpec
-from vllm.v1.outputs import (EMPTY_MODEL_RUNNER_OUTPUT, DraftTokenIds,
-                             ModelRunnerOutput)
+from vllm.v1.outputs import (EMPTY_MODEL_RUNNER_OUTPUT, AsyncModelRunnerOutput,
+                             DraftTokenIds, ModelRunnerOutput)
 from vllm.v1.worker.worker_base import WorkerBase
 
 from vllm_ascend.ascend_config import init_ascend_config
@@ -191,7 +191,7 @@ class NPUWorker(WorkerBase):
     def execute_model(
         self,
         scheduler_output: "SchedulerOutput",
-    ) -> Optional[ModelRunnerOutput]:
+    ) -> Optional[Union[ModelRunnerOutput, AsyncModelRunnerOutput]]:
         intermediate_tensors = None
         if not get_pp_group().is_first_rank:
             intermediate_tensors = IntermediateTensors(
@@ -220,7 +220,7 @@ class NPUWorker(WorkerBase):
             new_output.kv_connector_output = kv_connector_output
             return new_output
 
-        assert isinstance(output, ModelRunnerOutput)
+        assert isinstance(output, (ModelRunnerOutput, AsyncModelRunnerOutput))
         return output
 
     def load_model(self) -> None: