fix(desktop): persist transcript-only batch output

[ComputelessComputer]

Synthesize word timings for transcript-only batch responses and surface empty batch output as a no-speech error.

---

Note

Medium Risk
Moderate risk: changes core batch-transcription result handling and introduces synthesized timing data plus a new failure path when no words are produced.

Overview
Batch transcription now handles transcript-only outputs. When batch/streamed responses contain a transcript but no words, the listener synthesizes WordEntry timings (using segment start/duration or overall metadata duration) so results can still be persisted.

Empty batch output is now treated as an error. handleBatchResponse returns a boolean and runBatchSession rejects/marks the session failed with EMPTY_BATCH_TRANSCRIPT_ERROR when no words are produced; tests were added/updated to cover transcript-only persistence and the new no-speech failure behavior.

^{Reviewed by Cursor Bugbot for commit bab1216. Bugbot is set up for automated code reviews on this repo. Configure here.}

[netlify]

✅ Deploy Preview for old-char canceled.

Name	Link
🔨 Latest commit	bab1216
🔍 Latest deploy log	https://app.netlify.com/projects/old-char/deploys/6a0c5501fea4aa0009d067e2