feat: add WASAPI loopback capture mixed with microphone (Windows)

[MoulinLouis]

Before Submitting This PR

• I have searched existing issues and pull requests (including closed ones) to ensure this isn't a duplicate
• I have read CONTRIBUTING.md

Human Written Description

I do a lot of technical interviews over Google Meet and I've been wanting to transcribe both sides of the conversation, not just my mic. Right now I have to use a separate app for that and it's annoying. Since cpal already supports WASAPI loopback natively, I figured I'd try adding it to Handy.

I saw #248 and @cjpais's response about being open to a well-written PR. I also know about the feature freeze, so totally fine if this gets rejected.

I tried to keep it as non-intrusive as possible:

• It's Windows-only, uses WASAPI loopback which cpal already supports, so no new deps
• It mixes system audio with the mic instead of replacing it, so you still get both sides
• Off by default, and when it's off the code path is exactly the same as before
• One setting, two commands, one small toggle at the bottom of the Sound section

Related Issues/Discussions

Related to #248, same feature request but with a mic+loopback mixing approach instead of switching between sources.

Community Feedback

A few people asked for this in #248 for similar use cases (transcribing meetings, calls, etc). It's definitely a niche feature though, which is why I kept it as a simple opt-in toggle.

Testing

• Toggling "Record System Audio" on while playing a YouTube video: transcription captures the video audio mixed with mic input
• Toggling it off: normal microphone-only transcription works exactly as before
• Changing the output device while loopback is active restarts the stream correctly
• Mute-while-recording is correctly disabled when loopback is on (muting the output device would silence the capture source)
• Audio feedback (start/stop beeps) still plays in loopback mode
• Always-on microphone mode works correctly with loopback enabled
• Verified on Windows 11 with WASAPI

Changes overview

Backend (Rust):

• settings.rs: new record_system_audio: bool field
• recorder.rs: open() accepts optional loopback device, run_consumer() resamples both streams to 16kHz and mixes them before VAD
• audio.rs: always resolves mic device, optionally resolves output device for loopback
• commands/audio.rs: set_record_system_audio / get_record_system_audio commands
• actions.rs: mute skipped when loopback active (output device is the capture source)

Frontend (React/TypeScript):

• RecordSystemAudio.tsx: toggle component following existing patterns
• GeneralSettings.tsx: toggle at bottom of Sound section, MuteWhileRecording disabled when active
• settingsStore.ts: updater and normalization
• i18n translations added to all 20 locales

Screenshots/Videos (if applicable)

N/A, simple toggle in existing Sound settings section

AI Assistance

• AI was used (please describe below)

If AI was used:

• Tools used: Claude Code (claude-opus-4-6)
• How extensively: AI did most of the implementation, I handled the design decisions and reviewed everything

[cjpais]

Can you send screenshots, I'm open to this as a lot of people ask and if it's well implemented we can pull it in.

[MoulinLouis]