Empirically Revisiting and Enhancing IR-based Test-Case Prioritization