feat: enhance prompt selector with category tags and improve benchmark task processing

Add visual indicators for prompt categories and source types in the prompt selector component, and refactor benchmark task execution to use Dify streaming analysis with proper error handling and text extraction from Alibaba Cloud transcription results.

frontend/app/web-gold/src/views/system/task-management/layout/TaskLayout.vue

@@ -27,7 +27,7 @@
 
 <script setup>
 import { ref, computed, defineAsyncComponent, markRaw, onMounted, watch } from 'vue'
-import { VideoCameraOutlined, UserOutlined } from '@ant-design/icons-vue'
+import { VideoCameraOutlined, UserOutlined, FormOutlined } from '@ant-design/icons-vue'
 
 const STORAGE_KEY = 'task-management-active-tab'
 
@@ -50,6 +50,12 @@ const NAV_ITEMS = [
     label: '数字人视频任务',
     icon: UserOutlined,
     component: markRaw(defineAsyncComponent(() => import('../digital-human-task/index.vue')))
+  },
+  {
+    type: 'style-task',
+    label: '风格任务',
+    icon: FormOutlined,
+    component: markRaw(defineAsyncComponent(() => import('../../../task-center/BenchmarkTaskList.vue')))
   }
 ]