Function quantized_gemm_int8 

pub fn quantized_gemm_int8(
    a: &[i8],
    b: &[i8],
    m: usize,
    k: usize,
    n: usize,
    a_params: &QuantParams,
    b_params: &QuantParams,
) -> Vec<f32>

INT8 matrix multiply with f32 accumulation: C = A · B A: (m × k) as i8, B: (k × n) as i8, C: (m × n) as i32 → f32