Designing optimal experiments: an application to proton Compton scattering