Implement format conversion from ogg to mp3

[vansisto]

Telegram provides a voice in .ogg format. But OpenAI requires .mp3, or .wav (or other). .mp3 is the most appropriate format.

[vansisto]

I see, that I don't need to have a block for translation, as AI supports it by default. Enough send requests not to the "transcriptions" endpoint, but to "translations" instead.

[vansisto]

Wrong. It supports translations automatically for audio. But for test prompt it still translation has to be