Add support for Gemini API and Gemini TTS.

[YorkviskBR]

Gemini is one of the best LLM models out there today. Please add it.

[smirgol]

LLM != TTS - thus I recommend creating one request per feature, makes things easier for the maintainer. ;)

#217 does now implement the Google Gemini LLM.

[smirgol]

@YorkviskBR speaking of TTS, are you explicitly referring to the Gemini TTS, or just Google TTS in general?

The regular Google TTS is pretty straightforward to implement.
The Gemini TTS is rather new I guess, it looks very interesting and adds a nice touch, but might get expensive, too. :)
And the speed is noticeably slower than regular Google TTS.

FWIW:
We can probably use the Gemini TTS with some basic instructions. This worked well for me:

Style instructions: The text in brackets indicate the tone. Don't read the text in the brackets aloud!

[happy] "Glad that you are back at home!" [Bored] "it was sooo boring without you." [angry] "You really should have come earlier!"

Can be tested here: https://aistudio.google.com/generate-speech