Inconsistency of central measures in plot_direct_contribution_curves

[PabloRoque]

Channel contributions are aggregated using the mean of the Deterministic channel_contributions.

channel_contributions = self.compute_channel_contribution_original_scale().mean(
            ["chain", "draw"]
        )

We then overlay with the fitted params using the median:

def _plot_response_curve_fit(...):
...

    mid_params = self.format_recovered_transformation_parameters(quantile=0.5)
        y_fit = self.saturation.function(
            x=x_fit, **mid_params[channel]["saturation_params"]
        ).eval()

Easy solution:

• Establish the point estimates of channel_contributions using median.

Better solution:

• Enable a measure | estimand param in plot_direct_contribution_curves (possibly involving an Enum) that allows the user to select the measure of the point estimates.