Privacy docs now contradict the cloud ASR engine ("there is no transcription API call")

[Optic00]

With the OpenAI-compatible cloud ASR engine merged (#401), the published privacy docs are now materially wrong:

• docs/privacy/how-on-device-works.mdx still says transcription runs "entirely in-process -- ... there is no transcription API call" and "None of these requests include your audio, transcript, or summary content."
• docs/getting-started/what-is-steno.mdx still frames cloud access as summary-generation only.

When the user selects the Cloud API engine, audio is uploaded to the configured endpoint - the in-app confirm dialog says so correctly, but the docs are exactly what a privacy-sensitive user checks before trusting the app, and right now they deny the feature exists.

Needed: a section describing the opt-in cloud ASR path (what leaves the machine, when, to whom, that the key is stored encrypted via safeStorage and never written to config.json), plus softening the absolute claims to "with the default on-device engines". Should land before the next release ships the feature.

[Optic00]

The #410 revert removed the cloud ASR path entirely, so the privacy docs are accurate again and there is no disclosure gap to fix. Closing. If cloud ASR re-lands, the disclosure requirement should ship with it - the hardening notes on #403 include this.