[Core][Misc] Clean up ProfileExecuteDuration (#6461)

### What this PR does / why we need it?
This PR removes the custom `ProfileExecuteDuration` utility and its
usages across the codebase. This utility was used for profiling
execution duration of different stages in the inference process. It is
replaced by the standard `vllm.v1.utils.record_function_or_nullcontext`,
which integrates with PyTorch's profiler.

This change simplifies the code by removing a custom implementation in
favor of an upstream utility, improving maintainability. Associated
documentation and tests for `ProfileExecuteDuration` are also removed.

### Does this PR introduce _any_ user-facing change?
`VLLM_ASCEND_MODEL_EXECUTE_TIME_OBSERVE` env is removed now.

### How was this patch tested?
CI passed. The changes are a cleanup and replacement with a standard
utility. Existing tests cover the functionality. The removed feature had
its own tests which are also removed.

Related RFC: #5304

- vLLM version: v0.14.1
- vLLM main:
dc917cceb8

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

vllm_ascend/utils.py

@@ -23,7 +23,7 @@ import atexit
 import functools
 import math
 import os
-from contextlib import contextmanager, nullcontext
+from contextlib import nullcontext
 from enum import Enum
 from functools import lru_cache
 from threading import Lock
@@ -32,7 +32,6 @@ from typing import TYPE_CHECKING, Any
 import torch
 import torch_npu  # noqa: F401
 from packaging.version import InvalidVersion, Version
-from torch_npu.npu.streams import Event
 from vllm.logger import logger
 from vllm.sequence import IntermediateTensors
 
@@ -562,53 +561,6 @@ def dispose_tensor(x: torch.Tensor):
     x.set_(torch.empty((0,), device=x.device, dtype=x.dtype))
 
 
-class ProfileExecuteDuration:
-    _instance = None
-    _observations: list[tuple[str, Event, Event]] = []
-    _lock = Lock()
-
-    def __new__(cls):
-        with cls._lock:
-            if cls._instance is None:
-                cls._instance = super().__new__(cls)
-                atexit.register(cls._instance.destroy)
-            return cls._instance
-
-    def destroy(self):
-        with self._lock:
-            self._observations.clear()
-
-    @contextmanager
-    def capture_async(self, duration_tag: str):
-        if not envs_ascend.VLLM_ASCEND_MODEL_EXECUTE_TIME_OBSERVE:
-            yield
-            return
-
-        observe_start = Event(enable_timing=True)
-        observe_start.record()
-        try:
-            yield
-        finally:
-            observe_end = Event(enable_timing=True)
-            observe_end.record()
-            with self._lock:
-                self._observations.append((duration_tag, observe_start, observe_end))
-
-    def pop_captured_sync(self) -> dict:
-        """Pop and synchronize all events in the observation list"""
-        durations: dict[str, float] = {}
-        if not envs_ascend.VLLM_ASCEND_MODEL_EXECUTE_TIME_OBSERVE:
-            return durations
-
-        while self._observations:
-            with self._lock:
-                tag, observe_start, observe_end = self._observations.pop()
-            observe_end.synchronize()
-            durations[tag] = observe_start.elapsed_time(observe_end)
-
-        return durations
-
-
 def register_ascend_customop(vllm_config: VllmConfig | None = None):
     """Register Ascend CustomOP