A new Gradient TD Algorithm with only One Step-size: Convergence Rate Analysis using $L$-$\lambda$ Smoothness