Add quantized model support

[lazy-pr0grammer]

Hi, it's awesome that we can now run Whisper with php too. But normal models are very heavy and slow for CPU based architectures. Can you add the support for quantized models. They are much more faster