DOC remove confusing comment from TruncatedSVD

The comment was supposed to suggest that there might be a faster
way to do fit_transform than the matrix-vector multiplication
that it currently performs, but the suggested alternative actually
has higher complexity.

sklearn/decomposition/truncated_svd.py

@@ -120,7 +120,6 @@ def fit_transform(self, X, y=None):
         U, Sigma, VT = self._fit(X)
         Sigma = np.diag(Sigma)
 
-        # or (X * VT.T).T, whichever takes fewer operations...
         return np.dot(U, Sigma.T)
 
     def _fit(self, X):