refactor: 删除实时分析/按需识别/纯聊天三种模式切换

三种模式仅为前端 UI 区分，后端无感知，实际使用中价值不大：
- 纯聊天模式名不副实（VAD 仍附带摄像头帧）
- 实时分析增加 API 成本且体验不佳
- 按需识别已是默认且最自然的交互方式

删除后保留按需识别行为：语音带图 + 手动识别按钮，UI 更简洁。

frontend/src/lib/i18n/en-US.ts

@@ -48,7 +48,6 @@ export const enUS: TranslationMap = {
   "scenario.interpreter.hint": "AI will translate your speech in real-time (Chinese-English), without explanations",
 
   // Video indicators
-  "video.observing": "👁️ Observing",
   "video.listening": "🎤 Listening...",
   "video.playing": "🔊 Playing...",
   "video.initVad": "Initializing voice detection...",
@@ -67,8 +66,6 @@ export const enUS: TranslationMap = {
   "controls.cameraOn": "Turn on camera",
   "controls.micOff": "Turn off microphone",
   "controls.micOn": "Turn on microphone",
-  "controls.observing": "👁️ Observing",
-  "controls.observation": "👁️ Observe",
   "controls.interrupt": "⏹ Interrupt",
   "controls.stop": "End Session",
   "controls.stopVideo": "End Video",
@@ -77,7 +74,6 @@ export const enUS: TranslationMap = {
 
   // Chat panel
   "chat.title": "Chat",
-  "chat.mode.observation": "Observing",
   "chat.reconnecting": "Connection lost, reconnecting...",
   "chat.connecting": "Connecting to server...",
   "chat.vadInit": "Initializing voice detection...",
@@ -91,7 +87,6 @@ export const enUS: TranslationMap = {
   "chat.scenarioSwitched": "Switched to {name} mode",
 
   // Session messages
-  "session.changeDetected": "👁️ Scene change detected",
   "session.recognizing": "(Recognizing speech...)",
   "session.sttFailed": "(Speech recognition failed, please try again)",
   "session.noSpeech": "(No speech detected)",
@@ -141,9 +136,6 @@ export const enUS: TranslationMap = {
   "controls.device.camera": "Camera",
   "controls.device.mic": "Microphone",
   "controls.device.default": "Default",
-  "controls.mode.realtime": "Realtime",
-  "controls.mode.ondemand": "On-demand",
-  "controls.mode.chat": "Chat only",
 
   // Status bar
   "statusbar.ready": "Ready · Select devices to start",