Measuring the Algorithmic Convergence of Randomized Ensembles: The Regression Setting