feat(clients): event sound effects + optional text-to-speech

Add audible cues and optional spoken announcements for session events
(join/leave, channel + PM sent/recv, login, logout/connection-lost,
mic on/off, voice-activity, PTT) across all three clients, driven off
the shared C ABI vc_event stream so the mapping stays consistent.

TTS is off by default; when enabled it announces events and reads
message/PM bodies aloud. Master toggles + a sound-volume slider; the
per-utterance voice-activity and PTT cues default off. WAVs ship from
assets/sounds/.

Windows (built + verified): new VoiceCat.App/Notifications/ layer
(FeedbackSettings -> %AppData%\VoiceCat\feedback.json, SoundPlayerPool
via System.Media.SoundPlayer, SpeechAnnouncer via Prismatoid 0.3.0,
EventFeedback dispatcher); MainForm hooks; NotificationSettingsForm
under Settings > Notifications; csproj adds the Prismatoid PackageRef
and copies the WAVs into sounds\.

macOS + iOS (written, not yet built -- needs a Mac): shared
VoiceCatCore/Feedback/ (SoundEvent, EventFeedback = AVAudioPlayer pool
+ native AVSpeechSynthesizer, FeedbackSettings over UserDefaults); WAVs
bundled via Package.swift resources (.process). Hooks in SessionState/
AppState (iOS) and MainWindowController (macOS); settings UI in
SettingsView (iOS) and SettingsWindowController (macOS).

No core/server code touched; ctest --preset dev unaffected.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

clients/apple/Sources/VoiceCatCore/Feedback/EventFeedback.swift

@@ -0,0 +1,105 @@
+// EventFeedback — shared audible + spoken feedback for session events, used by both the macOS
+// (AppKit) and iOS (SwiftUI) clients. Mirrors the Windows client's EventFeedback policy
+// (clients/windows/.../Notifications/EventFeedback.cs): the platform event handlers decide WHAT
+// to play (they own the model/nickname/channel context); this type owns the "should I, and how"
+// policy plus the AVFoundation playback/synthesis.
+//
+// Sound effects use AVAudioPlayer; spoken announcements use the OS-native AVSpeechSynthesizer.
+//
+// NOTE (iOS): on the voice path the app runs a play-and-record AVAudioSession (VPIO). Playing
+// these cues / speech over that session can interact with the live call (ducking, route, or the
+// mute switch). The session category should allow mixing — verify on device. This is the most
+// likely place for platform bugs to surface.
+
+import Foundation
+import AVFoundation
+
+/// User preferences for event sounds and spoken feedback, backed by UserDefaults so the macOS
+/// and iOS settings screens and this player share one source of truth.
+public struct FeedbackSettings: Sendable {
+    public var sounds: Bool
+    public var speech: Bool
+    public var volume: Float
+    public var selfTalkSounds: Bool
+    public var pttSound: Bool
+
+    static let keySounds = "feedback.sounds"
+    static let keySpeech = "feedback.speech"
+    static let keyVolume = "feedback.volume"
+    static let keySelfTalk = "feedback.selfTalk"
+    static let keyPtt = "feedback.ptt"
+
+    /// Default values registered with UserDefaults (so "unset" reads as the intended default
+    /// rather than false/0).
+    static let defaults: [String: Any] = [
+        keySounds: true,
+        keySpeech: false,
+        keyVolume: 1.0,
+        keySelfTalk: false,
+        keyPtt: false,
+    ]
+
+    /// The current settings, read live from UserDefaults.
+    public static var current: FeedbackSettings {
+        let d = UserDefaults.standard
+        d.register(defaults: defaults)  // idempotent — ensures "unset" reads as the intended default
+        return FeedbackSettings(
+            sounds: d.bool(forKey: keySounds),
+            speech: d.bool(forKey: keySpeech),
+            volume: d.float(forKey: keyVolume),
+            selfTalkSounds: d.bool(forKey: keySelfTalk),
+            pttSound: d.bool(forKey: keyPtt))
+    }
+}
+
+@MainActor
+public final class EventFeedback {
+    public static let shared = EventFeedback()
+
+    private var players: [SoundEvent: AVAudioPlayer] = [:]
+    private let synthesizer = AVSpeechSynthesizer()
+
+    private init() {
+        UserDefaults.standard.register(defaults: FeedbackSettings.defaults)
+    }
+
+    // MARK: - Sounds
+
+    /// Play an event cue, honouring the user's settings. The two opt-in categories (your own
+    /// voice-activity, and the PTT cue) are gated by their own flags.
+    public func play(_ event: SoundEvent) {
+        let s = FeedbackSettings.current
+        guard s.sounds, s.volume > 0 else { return }
+        if (event == .vaStart || event == .vaStop), !s.selfTalkSounds { return }
+        if event == .ptt, !s.pttSound { return }
+
+        guard let player = player(for: event) else { return }
+        player.volume = s.volume
+        player.currentTime = 0
+        player.play()
+    }
+
+    /// Lazily load and cache an AVAudioPlayer for the event's bundled WAV. Returns nil (silent)
+    /// if the resource is missing or fails to load.
+    private func player(for event: SoundEvent) -> AVAudioPlayer? {
+        if let cached = players[event] { return cached }
+        guard let url = Bundle.module.url(forResource: event.resourceName, withExtension: "wav"),
+              let player = try? AVAudioPlayer(contentsOf: url) else {
+            return nil
+        }
+        player.prepareToPlay()
+        players[event] = player
+        return player
+    }
+
+    // MARK: - Speech
+
+    /// Speak `text` when spoken feedback is enabled. Utterances queue (do not interrupt prior
+    /// speech) so a burst of events is read in order.
+    public func speak(_ text: String) {
+        guard FeedbackSettings.current.speech else { return }
+        let trimmed = text.trimmingCharacters(in: .whitespacesAndNewlines)
+        guard !trimmed.isEmpty else { return }
+        synthesizer.speak(AVSpeechUtterance(string: trimmed))
+    }
+}

clients/apple/Sources/VoiceCatCore/Feedback/SoundEvent.swift

@@ -0,0 +1,43 @@
+// SoundEvent — the cross-platform set of audible event cues. Each maps to a WAV bundled as an
+// SPM resource (Sources/VoiceCatCore/Sounds/, copied from assets/sounds/). The same logical set
+// is mirrored in the Windows client (clients/windows/.../Notifications/SoundEvent.cs) so feedback
+// stays consistent across platforms.
+
+import Foundation
+
+public enum SoundEvent: CaseIterable, Sendable {
+    case channelJoin    // another user joined my channel
+    case channelLeave   // another user left my channel
+    case channelRecv    // channel text message from someone else
+    case channelSent    // channel text message I sent
+    case pmRecv         // private message received
+    case pmSent         // private message I sent
+    case login          // connected / authenticated
+    case logout         // clean disconnect
+    case connectionLost // unexpected disconnect
+    case voiceOn        // my microphone stream started
+    case voiceOff       // my microphone stream stopped
+    case vaStart        // my voice-activity began (off by default)
+    case vaStop         // my voice-activity ended (off by default)
+    case ptt            // push-to-talk engaged (off by default)
+
+    /// Resource name (without extension) as bundled in Sources/VoiceCatCore/Sounds/.
+    var resourceName: String {
+        switch self {
+        case .channelJoin:    return "channel_join"
+        case .channelLeave:   return "channel_leave"
+        case .channelRecv:    return "channel_recv"
+        case .channelSent:    return "channel_sent"
+        case .pmRecv:         return "pm_recv"
+        case .pmSent:         return "pm_sent"
+        case .login:          return "login"
+        case .logout:         return "logout"
+        case .connectionLost: return "connection_lost"
+        case .voiceOn:        return "voice_on"
+        case .voiceOff:       return "voice_off"
+        case .vaStart:        return "va_start"
+        case .vaStop:         return "va_stop"
+        case .ptt:            return "ptt"
+        }
+    }
+}