Context-dependent Instruction Tuning for Dialogue Response Generation