Function layer_norm_slow

pub fn layer_norm_slow(
    x: &Tensor,
    alpha: &Tensor,
    beta: &Tensor,
    eps: f32,
) -> Result<Tensor>