feat(cli): speech-to-text v2 — streaming

[claude]

Adds CLI recipe for live streaming transcription using Speech-to-Text v2 API with flux-general-en model.

Resolves #201

[github-actions]

✅ CLI Recipe Tests

── recipes/cli/speech-to-text/v2/streaming ──
PASS
/home/runner/work/recipes/recipes/recipes/cli/speech-to-text/v2/streaming/example.sh: line 13: ffmpeg: command not found
✓ Streaming stdin · nova-3 · send EOF (Ctrl+D) to finish
✓ PASSED

[claude]

@deepgram-devrel — VP escalation: PR #212 has been stuck for >2 days with failing CI checks and re-trigger attempts have not resolved it. Manual review needed.

State: labels=[type:samples, language:cli], checks: test-cli=FAILURE, e2e-api-check=FAILURE. No lead-fix workflow available to auto-remediate.

[lukeocodes]

Closing — CLI v2 streaming is not supported. The Deepgram CLI only routes flux-general-en to the /v2 endpoint for URL-based (pre-recorded) transcription, not for stdin pipe streaming. v2 pre-recorded is covered by PR #214. v2 WebSocket streaming requires the SDK directly, not the CLI.