Chang SGDClassifier loss from 'log' to 'log_loss' (#9113)

scikit-learn/scikit-learn/pull/23046 Introduces `'log_loss'` in
`SGDClassifier` and deprecates `'log'`. This PR addresses this issue in
the repository (no other files in the repo need to be modified for this
issue)

examples/graph_sage_unsup_ppi.py

@@ -79,7 +79,7 @@ def test():
     # Train classifier on training set:
     x, y = encode(train_loader)
 
-    clf = MultiOutputClassifier(SGDClassifier(loss='log', penalty='l2'))
+    clf = MultiOutputClassifier(SGDClassifier(loss='log_loss', penalty='l2'))
     clf.fit(x, y)
 
     train_f1 = f1_score(y, clf.predict(x), average='micro')