Function dequant_nvfp4_to_bf16 

pub fn dequant_nvfp4_to_bf16(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    packed: DevicePtr,
    scales: DevicePtr,
    out_bf16: DevicePtr,
    scale2: f32,
    n: u32,
    k: u32,
    stream: u64,
) -> Result<()>

2026-09-25: Dequantize an NVFP4 weight [n, k] (packed E2M1, E4M3 group scales and the per-tensor scale2) to BF16 [n, k], one block per row.