Desktop: dictate on this Mac with Apple Speech (on-device, like iOS)

[RVZO6]

Problem

On iOS, T3 Code dictates with Apple's on-device SpeechAnalyzer. On macOS, the desktop app can only transcribe through an environment model (which you must download) or a cloud provider. A Mac on macOS 26 or later has the same Apple speech engine as the iPhone, but T3 Code cannot use it.

Proposal

Add On this Mac → Apple Speech as a per-device destination in Settings → Voice → Transcribe with, in the macOS desktop app only.

• It is a destination of the device, not an environment model. Environment model lists, the server and mobile do not change. Remote environments and other clients never see it, so there is no "model you cannot use remotely" problem.
• A small Swift CLI helper, adapted from the mobile Apple transcription code, runs prepare (resolves the language and installs Apple's assets) and transcribe <wav>. The Electron main process calls it through IPC. Audio stays on the Mac.
• The macOS release runners already have Xcode 26, so they build the helper as a universal binary and ship it in Resources. Dev builds without Xcode 26 skip it, and the option does not show.
• Not in scope: web, Windows/Linux, and hosting Apple Speech for other devices. Silence filtering, filler-word removal and custom words still apply only to environment and cloud transcription.

Related: #24938.

Draft implementation for review of the direction: #34027.