`!tts`: can't generate longer than 10s audio without generating bad audio

In the meantime, I am setting [this](https://github.com/sudomaze/OnlySudo/blob/main/src/ai/tts/main.py#L61) to read the entire message, but this causes the final audio to be scuffed. Need to find a way to generate longer audio while keeping the quality of the model.