Update dim-reduction-overview.md

Why we should do standardization of the data? I think the centering is included for evaluating the covariance matrix when we use sklearn-pca, and the same variances in the features pervertes the outcome

in kernel-pca we do not use Standardscaler()