Groq as a cloud dictation provider

[varo6]

Now that cloud dictation is in (#31769), I'd like to suggest adding Groq as a provider. Groq with a q, the inference company, not xAI's Grok.

The main reason is speed. Groq runs Whisper Large v3 Turbo on their own hardware and transcription comes back almost instantly, which matters a lot for dictation. That's also why apps like FreeFlow on Mac use it by default. On top of that the free tier is generous enough for personal dictation, and past that it's $0.04 per hour of audio. Inference requests aren't retained by default and ZDR can be enabled from the account settings, so it should fit the current privacy model.

You can technically reach Whisper on Groq through OpenRouter already, but OpenRouter splits that model between DeepInfra and Groq, so you don't reliably get Groq's latency, and you can't use your own Groq key or its free tier.

The API is OpenAI-compatible (/openai/v1/audio/transcriptions), so it looks like it would follow the same path as the OpenAI and xAI providers in CloudSpeech.ts, with custom words sent through prompt.

If you're open to it, I'd be happy to send a small PR. Thanks!