indpependence assumption is more restrictive than zero correlation. indppendence implies martingale difference which further implies zero correlation. However, for the variance ratio test, zero corrlation is enough, becasue it suffices to show that the so-called "long run variance" is equal to "short run variance", i.e., dependence structure plays no role in the statistics. In fact, the variance of the partial sum of zero correlation process increase linearly which forms the basis of VR test.
zero correlation doesn't imply independence. A simple example, X has symmetric distribution, of course, X and X^(2) are not independent, but they are uncorrelated. Because E(X.X^2)=E(X^3)=0. Coversely, independence implies zero correlation.