Support for Int8 quantization

For now, the PTQ pipeline applies fake quantization to int8, but casting options are limited to 32 bits. The goal is to add support for INT8 casting.