feat(trace): 在请求链路追踪中记录并展示首字时间(TTFB)

将 first_byte_time_ms 添加到候选记录的 extra_data 中，
并在前端时间线组件中展示该指标。

frontend/src/features/usage/components/HorizontalRequestTimeline.vue

@@ -175,6 +175,13 @@
                       {{ currentAttempt.finished_at ? formatTime(currentAttempt.finished_at) : '进行中' }}
                     </span>
                   </div>
+                  <div
+                    v-if="currentAttempt.extra_data?.first_byte_time_ms != null"
+                    class="info-item"
+                  >
+                    <span class="info-label">首字 (TTFB)</span>
+                    <span class="info-value mono">{{ formatLatency(currentAttempt.extra_data.first_byte_time_ms) }}</span>
+                  </div>
                   <div
                     v-if="currentAttempt.key_name || currentAttempt.key_id"
                     class="info-item"

src/api/handlers/base/cli_handler_base.py

@@ -1438,6 +1438,13 @@ class CliMessageHandlerBase(BaseMessageHandler):
                     if ctx.status_code and ctx.status_code >= 400:
                         # 请求链路追踪使用 upstream_response（原始响应），回退到 error_message（友好消息）
                         trace_error_message = ctx.upstream_response or ctx.error_message or f"HTTP {ctx.status_code}"
+                        extra_data = {
+                            "stream_completed": False,
+                            "chunk_count": ctx.chunk_count,
+                            "data_count": ctx.data_count,
+                        }
+                        if ctx.first_byte_time_ms is not None:
+                            extra_data["first_byte_time_ms"] = ctx.first_byte_time_ms
                         RequestCandidateService.mark_candidate_failed(
                             db=bg_db,
                             candidate_id=ctx.attempt_id,
@@ -1447,23 +1454,22 @@ class CliMessageHandlerBase(BaseMessageHandler):
                             error_message=trace_error_message,
                             status_code=ctx.status_code,
                             latency_ms=response_time_ms,
-                            extra_data={
+                            extra_data=extra_data,
-                                "stream_completed": False,
-                                "chunk_count": ctx.chunk_count,
-                                "data_count": ctx.data_count,
-                            },
                         )
                     else:
+                        extra_data = {
+                            "stream_completed": True,
+                            "chunk_count": ctx.chunk_count,
+                            "data_count": ctx.data_count,
+                        }
+                        if ctx.first_byte_time_ms is not None:
+                            extra_data["first_byte_time_ms"] = ctx.first_byte_time_ms
                         RequestCandidateService.mark_candidate_success(
                             db=bg_db,
                             candidate_id=ctx.attempt_id,
                             status_code=ctx.status_code,
                             latency_ms=response_time_ms,
-                            extra_data={
+                            extra_data=extra_data,
-                                "stream_completed": True,
-                                "chunk_count": ctx.chunk_count,
-                                "data_count": ctx.data_count,
-                            },
                         )
 
             finally:

src/api/handlers/base/stream_telemetry.py

@@ -241,16 +241,20 @@ class StreamTelemetryRecorder:
 
         from src.services.request.candidate import RequestCandidateService
 
+        extra_data: Dict[str, Any] = {
+            "stream_completed": ctx.is_success(),
+            "data_count": ctx.data_count,
+        }
+        if ctx.first_byte_time_ms is not None:
+            extra_data["first_byte_time_ms"] = ctx.first_byte_time_ms
+
         if ctx.is_success():
             RequestCandidateService.mark_candidate_success(
                 db=db,
                 candidate_id=ctx.attempt_id,
                 status_code=ctx.status_code,
                 latency_ms=response_time_ms,
-                extra_data={
+                extra_data=extra_data,
-                    "stream_completed": True,
-                    "data_count": ctx.data_count,
-                },
             )
         else:
             error_type = "client_disconnected" if ctx.status_code == 499 else "stream_error"
@@ -263,10 +267,7 @@ class StreamTelemetryRecorder:
                 error_message=trace_error_message,
                 status_code=ctx.status_code,
                 latency_ms=response_time_ms,
-                extra_data={
+                extra_data=extra_data,
-                    "stream_completed": False,
-                    "data_count": ctx.data_count,
-                },
             )
 
     async def _update_usage_status_on_error(