Computes the dependencies (as measures by mutual information) between all
positions (columns) of discrete data. Specifically, it returns for each pair
of positions (i,j) the mutual information I(X_i,X_j) multiplied by the number
N of sequences (rows), which may also be used for testing the statistical
significance of mutual information values, as for large N, 2*N*I(X_i,X_j) is
approximately chi squared.