C3-STISR: Scene Text Image Super-resolution with Triple Clues