Sub-issue: live transcription pipeline

## What
Add a live / incremental transcription flow so EchoNote shows text while recording instead of waiting until the end.

## Why
This is the biggest gap between EchoNote and the Meetily-style experience. It makes the app feel responsive and useful during long calls.

## How
- stream audio chunks during recording
- feed chunks into the local Whisper pipeline incrementally
- persist partial transcript updates to SQLite
- update the UI with live transcript progress
- keep the final transcript compatible with the current summary flow

Parent: #51

Start by tracing the recording flow and locating the local Whisper pipeline, SQLite transcript persistence, and transcript UI entry points. Implement incremental audio processing and partial updates while preserving the final transcript consumed by the existing summary flow; done means text appears during recording and final summaries remain compatible.

Written by the indexing model from the issue text.