For some reason, doing the interpolation for the exact recipe, in

    def evaluate_theory_prediction_lensing_profile()
    .
    .
    .
            if self.cluster_theory._beta_parameters is not None:
                self.cluster_theory.set_beta_s_interp(*z_edges)

is not giving good results. Also since we expect the cluster_theory has to be prepared with the beta parameters, I think it also should be prepared with the right beta interpolation, the same way we do for the grid recipe.

To see the error, just run the cluster_recipe.ipynb notebook in #63.
Without this fix, the figure in Cell 7, Reduced Shear — Exact vs Grid (No Interp) for 3 Radii shows that the prediction do not agree, even though both are prepared with the same cluster_theory object.