Provide a tutorial page to describe all tricks to improve GraphStorm training/inference speed

There are multiple ways of speeding up GraphStorm training and inference:
- [ ] create EC2 instances in the same placement group
- [ ] use WholeGraph

Start by locating GraphStorm's existing tutorial and documentation entry points and reviewing how training and inference guidance is organized. The tutorial should explain the listed EC2 placement-group and WholeGraph techniques, along with the other speed improvements identified during research, and make clear how readers can apply them.

Written by the indexing model from the issue text.