feat: ASR语音转写管线 + 群聊身份混淆修复

- 新增ASR语音识别管线: QQ语音→下载音频→qwen3-asr-flash转录→注入用户消息
- 模型名称全部从models.json路由获取，无硬编码
- 修复群聊中AI将非管理员用户误称为管理员昵称(叶酱)的问题
  - 助手回复缓存时标注[回复 昵称 (UID)]，防止对话历史中身份混淆
  - 群聊上下文指令改为肯定性表述，移除具体名称提及
- trace面板时间戳改为YYYY-MM-DD HH:MM:SS格式，耗时统一显示为秒
- 修复Go time.Duration纳秒值在前端显示问题(Duration/1e6转毫秒)
- 新增video_tool插件模板
- 优化OpenAI adapter reasoning_content处理

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

backend/ai-core/internal/llm/selector.go

@@ -18,8 +18,10 @@ const (
 	PurposeIntentAnalysis   ModelPurpose = "intent_analysis"
 	PurposeToolCalling      ModelPurpose = "tool_calling"
 	PurposeMemoryExtraction ModelPurpose = "memory_extraction"
-	PurposeVision           ModelPurpose = "vision"
-	PurposeOCR              ModelPurpose = "ocr"
+	PurposeVision              ModelPurpose = "vision"
+	PurposeVideo               ModelPurpose = "video"
+	PurposeOCR                 ModelPurpose = "ocr"
+	PurposeSpeechRecognition   ModelPurpose = "speech_recognition"
 )
 
 // ErrModelNotRequired is returned when an optional model is unavailable.