Generalising unit-refutation completeness and SLUR via nested input resolution