pub fn fp4_quant_e4m3_inplace(x: &mut [f32], block: usize)
fp4_act_quant(x, 16, inplace=True, scale_dtype=e4m3): amax floored at 6 * 2^-9, scale = amax / 6 rounded through e4m3, values to e2m1, dequantised back to bf16.
fp4_act_quant(x, 16, inplace=True, scale_dtype=e4m3)
6 * 2^-9
amax / 6