A Convergence Result for Non-Autonomous Subgradient Evolution Equations and Its Application to the Steepest Descent Exponential Penality Trajectory in Linear Programming