Deep copy needed before each scaling in predict_subset() method of pvops/timeseries/models/AIT.py

[agmoore4]

Currently, when new standardized columns are created, the original columns are also standardized. This results in the R2 and logRMSE being calculated between the predicted unscaled energy and the true scaled energy, suggesting (incorrectly) that the model performs poorly.

There is a deep copy at the start of the method, but it doesn't do anything. The deep copy needs to happen inside the for-loop before each standardization is applied.