Function per_token_group_quant_fp8 

pub fn per_token_group_quant_fp8(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    input_bf16: DevicePtr,
    output_fp8: DevicePtr,
    a_scale: DevicePtr,
    m: u32,
    k: u32,
    stream: u64,
) -> Result<()>

Per-token-per-128-K-group FP8 activation quantization. Output: A_fp8 [M, K] FP8 E4M3 + a_scale [M, K/128] FP32. Matches vLLM’s per_token_group_quant_fp8.

Grid: (K/128, M, 1) Block: (128, 1, 1)