Fix data normalization methods

[zimenglyu]

Describe your changes

Fix the data normalization methods so that data normalization bounds come from training data only

What type is this change?

• Bug fix
• New feature

Checklist before requesting a review

• My code follows the style guidelines of this project
• I have tested on MAC OS system
• I have tested on Ubuntu system
• I have tested on the RIT research cluster and updated the spack packages for compiling
• My changes generate no new warnings
• If the PR is related to a new feature, I added the running scripts for the new feature

How Has This Been Tested?

Added this log information for sanity check, the normalization bounds should only come from testing files:
Log::debug("Training file: %d, min: %lf, max: %lf\n", train_index, current_min, current_max);