Add VAD-based real-time speech recognition feedback

- Add SherpaOnnxVAD for voice activity detection with Silero model
- Implement streaming audio capture using AVAudioEngine
- Show real-time recognition results in overlay HUD while speaking
- Add intelligent text merging to handle segment overlaps
- Add VAD model auto-download support
- Bump version to 1.2.4

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

Author: ZhaoChaoqun

Sources/OverlayWindow.swift

@@ -17,7 +17,7 @@ class OverlayWindowController {
 
         // 创建一个无边框的悬浮窗口
         window = NSWindow(
-            contentRect: NSRect(x: 0, y: 0, width: 200, height: 36),
+            contentRect: NSRect(x: 0, y: 0, width: 320, height: 80),
             styleMask: [.borderless],
             backing: .buffered,
             defer: false
@@ -54,6 +54,7 @@ class OverlayWindowController {
     func show() {
         print(">>> OverlayWindow show() called")
         viewModel.state = .recording
+        viewModel.reset()
         viewModel.startAnimation()
         positionWindow()
         window?.orderFront(nil)
@@ -66,6 +67,14 @@ class OverlayWindowController {
         positionWindow()
     }
 
+    func updateRecognizedText(_ text: String) {
+        viewModel.recognizedText = text
+        // 窗口内容变化时重新定位
+        DispatchQueue.main.async { [weak self] in
+            self?.positionWindow()
+        }
+    }
+
     func hide() {
         print(">>> OverlayWindow hide() called")
         viewModel.stopAnimation()
@@ -82,6 +91,7 @@ class OverlayViewModel: ObservableObject {
 
     @Published var state: State = .recording
     @Published var animationPhase: CGFloat = 0
+    @Published var recognizedText: String = ""
 
     private var animationTimer: Timer?
 
@@ -96,47 +106,66 @@ class OverlayViewModel: ObservableObject {
         animationTimer?.invalidate()
         animationTimer = nil
     }
+
+    func reset() {
+        recognizedText = ""
+    }
 }
 
 /// 悬浮窗口视图 - 类似截图中的样式
 struct OverlayView: View {
     @ObservedObject var viewModel: OverlayViewModel
 
     var body: some View {
-        HStack(spacing: 12) {
-            // 左侧图标区域 - 固定宽度
-            Group {
-                if viewModel.state == .recording {
-                    // 录音动画 - 竖纹声波
-                    HStack(spacing: 3) {
-                        ForEach(0..<7, id: \.self) { index in
-                            RoundedRectangle(cornerRadius: 1.5)
-                                .fill(Color.white.opacity(0.9))
-                                .frame(width: 3, height: waveHeight(for: index))
+        VStack(spacing: 8) {
+            // 主状态指示器
+            HStack(spacing: 12) {
+                // 左侧图标区域 - 固定宽度
+                Group {
+                    if viewModel.state == .recording {
+                        // 录音动画 - 竖纹声波
+                        HStack(spacing: 3) {
+                            ForEach(0..<7, id: \.self) { index in
+                                RoundedRectangle(cornerRadius: 1.5)
+                                    .fill(Color.white.opacity(0.9))
+                                    .frame(width: 3, height: waveHeight(for: index))
+                            }
                         }
+                    } else {
+                        // 处理中 - 显示旋转指示器
+                        ProgressView()
+                            .progressViewStyle(CircularProgressViewStyle(tint: .white))
+                            .scaleEffect(0.7)
                     }
-                } else {
-                    // 处理中 - 显示旋转指示器
-                    ProgressView()
-                        .progressViewStyle(CircularProgressViewStyle(tint: .white))
-                        .scaleEffect(0.7)
                 }
+                .frame(width: 30, height: 18)
+
+                // 右侧文字区域 - 固定宽度确保一致
+                Text(viewModel.state == .recording ? "正在聆听..." : "识别中...")
+                    .font(.system(size: 13, weight: .medium))
+                    .foregroundColor(.white)
+                    .frame(width: 75, alignment: .leading)
             }
-            .frame(width: 30, height: 18)
 
-            // 右侧文字区域 - 固定宽度确保一致
-            Text(viewModel.state == .recording ? "正在聆听..." : "识别中...")
-                .font(.system(size: 13, weight: .medium))
-                .foregroundColor(.white)
-                .frame(width: 75, alignment: .leading)
+            // 识别结果显示（当有文字时）
+            if !viewModel.recognizedText.isEmpty {
+                Text(viewModel.recognizedText)
+                    .font(.system(size: 13))
+                    .foregroundColor(.white.opacity(0.9))
+                    .lineLimit(2)
+                    .truncationMode(.head)
+                    .frame(maxWidth: 280)
+                    .multilineTextAlignment(.center)
+            }
         }
         .padding(.horizontal, 16)
-        .padding(.vertical, 8)
+        .padding(.vertical, 10)
         .background(
             Capsule()
                 .fill(Color.black.opacity(0.75))
         )
         .animation(.easeInOut(duration: 0.1), value: viewModel.animationPhase)
+        .animation(.easeInOut(duration: 0.2), value: viewModel.recognizedText)
     }
 
     private func waveHeight(for index: Int) -> CGFloat {