Accessing pion GPDs through the Sullivan process: is it feasible?