Finite Sample Breakdown of $M$- and $P$-Estimators