K±p elastic scattering at 5 GeV/c. Evidence for a K−p backward peak