Numerical approximation of optimal control problems with $L^q$-quasinorm penalization ($q\in(0,1)$)