Add support for new giga asr models

[mikeozornin]

cjpais/Handy#1688 (comment)

Recently two ASR models were published:
https://huggingface.co/ai-sage/GigaAM-Multilingual
https://huggingface.co/ai-sage/GigaChat3.1-Audio-10B-A1.8B

In Russian, Kazakh, Kyrgyz, and Uzbek tests they performed better than whisper large. It will be nice to have it in Handy, maybe it's worth to replace older GigaAM models.

[cjpais]

I think GigaAM-Multilingual will probably come in first mainly because it's probably simpler to port and runs on more systems

[ekhodzitsky]

there's already an int8 onnx export of the multilingual model on hf: istupakov/gigaam-multilingual-ctc-onnx (~225mb). it's encoder-only ctc, decode is just argmax + collapse — much simpler than the rnnt heads. the 25s thing only happens if you run the encoder once on the whole file, split on silence and it's fine for hours