Approximate saddle-point theorems in vector optimization