Bugfix E-step math error

Commit 41f1090a was based around a mistaken claim that
    log(det(X)) = sum(log(diag(chol(X))))
but actually this is missing a factor of 2 because the Cholesky
decomposition is a sort of square root. So in reality it should be
    log(det(X)) = 2 * sum(log(diag(chol(X))))
and this commit incorporates that factor of 2.

@MoDT/Estep.m

@@ -40,7 +40,7 @@
     % Mahalanobis distance is cached from the M-step
     delta = self.mahal_dist;
     for k = 1:K
-        logSqrtDetC(k) = sum(log(diag(chol(self.C(:,:,k))))) / 2;
+        logSqrtDetC(k) = sum(log(diag(chol(self.C(:,:,k)))));
     end
 end