feat(ui): add floating caption window with live transcription

Add a frameless always-on-top caption window wired to the live capture
pipeline. Silence-based segmentation gates the audio so the model never
sees silence, and the panel background is painted in paintEvent with a
configurable opacity instead of a style sheet.

新增悬浮字幕窗（无边框、置顶、可折叠/展开），接入实时采集与流式转写。
分段由静音检测驱动，静音期间不送音频，避免模型对静音产生幻觉。底板
改为在 paintEvent 中绘制，不透明度可调。

src/livetranscriber.h

@@ -0,0 +1,52 @@
+#pragma once
+
+#include <QObject>
+#include <QString>
+
+#include <condition_variable>
+#include <mutex>
+#include <string>
+#include <thread>
+#include <vector>
+
+#include "audio.h"
+#include "streamingasr.h"
+
+namespace lc {
+
+class Qwen3Asr;
+
+// 实时字幕：采集默认输出 → 流式转写 → 用信号把文本推给 UI。
+// 分段由静音检测驱动：句子之间的自然停顿就收束一段、清空上下文，
+// 这样解码器的上下文不会无限增长（见 docs/phase2-streaming.md）。
+class LiveTranscriber : public QObject {
+    Q_OBJECT
+
+public:
+    explicit LiveTranscriber(Qwen3Asr & asr, QObject * parent = nullptr);
+    ~LiveTranscriber() override;
+
+    bool start(const StreamOptions & opts, QString * error = nullptr);
+    void stop();
+
+    QString deviceName() const;
+
+Q_SIGNALS:
+    // committed = 已稳定文本；tail = 不稳定尾部（UI 用暗色显示）
+    void caption(const QString & committed, const QString & tail, const QString & language);
+    void failed(const QString & message);
+
+private:
+    void workerLoop();
+
+    Qwen3Asr &             asr_;
+    MonitorCapture         capture_;
+    StreamOptions          opts_;
+    std::thread            worker_;
+    std::mutex             mutex_;
+    std::condition_variable cv_;
+    std::vector<float>     pending_;
+    bool                   quit_ = false;
+};
+
+} // namespace lc