Add local Whisper Telegram transcription

[jackgladowsky]

Summary

• add Telegram voice/audio/audio-document ingestion that transcribes locally with whisper.cpp
• add STT config for local whisper binary/model, ffmpeg conversion, size cap, and timeout
• feed transcripts into the normal chat handler with a transcribed prompt prefix
• document local whisper.cpp setup and add mocked deterministic tests

Tests

• git diff --check
• pnpm run typecheck
• pnpm run build
• pnpm test

Notes

• No OpenAI/API transcription required.
• Before deploy, live ~/.jarvis/config.yaml needs the new stt section and local whisper.cpp/model paths.