[Dataset] Fix the edge numbers in the GNNBenchmarkDataset doc string (#3163)

* PPIDataset

* Revert "PPIDataset"

This reverts commit 264bd0c960cfa698a7bb946dad132bf52c2d0c8a.

* Hide implementations that may cause confusions to the user

* fix number of classes in Amazon Cobuy datasets

* add four papers

* Fix doc string

* fix readme

* fix readme
Co-authored-by: Quan (Andy) Gan <coin2028@hotmail.com>

python/dgl/data/gnn_benchmark.py

@@ -219,7 +219,8 @@ class CoauthorCSDataset(GNNBenchmarkDataset):
     Statistics:
 
     - Nodes: 18,333
-    - Edges: 327,576
+    - Edges: 163,788 (note that the original dataset has 81,894 edges but DGL adds
+      the reverse edges and remove the duplicates, hence with a different number)
     - Number of classes: 15
     - Node feature size: 6,805
 
@@ -285,7 +286,8 @@ class CoauthorPhysicsDataset(GNNBenchmarkDataset):
     Statistics
 
     - Nodes: 34,493
-    - Edges: 991,848
+    - Edges: 495,924 (note that the original dataset has 247,962 edges but DGL adds
+      the reverse edges and remove the duplicates, hence with a different number)
     - Number of classes: 5
     - Node feature size: 8,415
 
@@ -350,7 +352,8 @@ class AmazonCoBuyComputerDataset(GNNBenchmarkDataset):
     Statistics:
 
     - Nodes: 13,752
-    - Edges: 574,418
+    - Edges: 491,722 (note that the original dataset has 245,778 edges but DGL adds
+      the reverse edges and remove the duplicates, hence with a different number)
     - Number of classes: 10
     - Node feature size: 767
 
@@ -415,7 +418,8 @@ class AmazonCoBuyPhotoDataset(GNNBenchmarkDataset):
     Statistics
 
     - Nodes: 7,650
-    - Edges: 287,326
+    - Edges: 238,163 (note that the original dataset has 119,043 edges but DGL adds
+      the reverse edges and remove the duplicates, hence with a different number)
     - Number of classes: 8
     - Node feature size: 745