Quantization trait: Directly give output scaling factors

[nikkolasg]

Right now, the framework forces every layer to do the following to get the corresponding output scaling factors:

fn quantize_op<S: ScalingStrategy>(
        self,
        data: &S::AuxData,
        node_id: NodeId,
        input_scaling: &[ScalingFactor],
    ) -> anyhow::Result<QuantizeOutput<Self::QuantizedOp>> {
        let mut output_scalings = S::scaling_factors_for_node(data, node_id, 1);
        ensure!(
            output_scalings.len() == 1,
            "Output scaling for convolution layer different from 1"
        );
     ...

The process could be streamlined if we directly give the output scaling factors as well, since the trait ScalingStrategy can only be used for that anyway.

[linear]

CRY-66 Quantization trait: Directly give output scaling factors