Add note for edge_size coefficient in GNNExplainer (#8215)

Following from #1985

The default value of edge_size coefficient is suitable for many
networks, but at the end it is tightly related with the number of edges.
It is useful to reevaluate if we want to use this number based on our
objective of "compactness" of the explanation.

I consider that this value is not properly announced (my feeling is that
normally it is just ignored) and with this PR I try to raise awareness
about it with a note in the comments of the class.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: rusty1s <[email protected]>

torch_geometric/explain/algorithm/gnn_explainer.py

@@ -29,6 +29,18 @@ class GNNExplainer(ExplainerAlgorithm):
         gnn_explainer_link_pred.py <https://github.com/pyg-team/
         pytorch_geometric/blob/master/examples/explain/gnn_explainer_link_pred.py>`_.
 
+    .. note::
+
+        The :obj:`edge_size` coefficient is multiplied by the number of nodes
+        in the explanation at every iteration, and the resulting value is added
+        to the loss as a regularization term, with the goal of producing
+        compact explanations.
+        A higher value will push the algorithm towards explanations with less
+        elements.
+        Consider adjusting the :obj:`edge_size` coefficient according to the
+        average node degree in the dataset, especially if this value is bigger
+        than in the datasets used in the original paper.
+
     Args:
         epochs (int, optional): The number of epochs to train.
             (default: :obj:`100`)