Universal stabilization using control Lyapunov functions, adaptive derivative feedback, and neural network approximators