Fix GNNExplainer with GCN

[RBendias]

Currently, GNNExplainer uses the wrong number of hops in the GNNExplainer examples. The default value is equal to the number of message passing layers, but for an n-layer GCN model, the number of hops is equal to n+1.
Increasing num_hops by 1 would include too many edges in the explanation.
Instead, I excluded the normalization from each GCNConv and added GCNNorm as a data transformation.

[codecov-commenter]

Codecov Report

Merging #3508 (90c94f3) into master (83d929c) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #3508   +/-   ##
=======================================
  Coverage   81.70%   81.70%           
=======================================
  Files         291      291           
  Lines       14465    14465           
=======================================
  Hits        11818    11818           
  Misses       2647     2647           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 83d929c...90c94f3. Read the comment docs.

[rusty1s]

LGTM. Thank you.