Implement PCA loadings

[jediknight6]

Add a pca_loadings() function that returns the contribution of each original feature to each principal component.

PCA is an unsupervised dimensionality-reduction technique. While explained variance shows how much variance each component captures, PCA loadings help explain which original columns contribute most to those components. This would make Percentify’s PCA utilities more interpretable and useful for exploratory data analysis before model training.

[OhhJJ]

Would it be worth clarifying whether pca_loadings() should return the raw component weights (components_.T) or the scaled loadings (components_.T * sqrt(explained_variance_))? Both definitions are commonly used, so it would help avoid ambiguity during implementation.

[theloyaldaniil-tech]

It might be helpful to specify the expected return type (e.g. a NumPy array or a pandas DataFrame), especially if preserving feature names is desired.

[andreyfatey10]

It would also be nice to include a short example of the expected output in the issue description or documentation to make the intended behavior clearer.

[Sanpen43]

It would be helpful to specify whether these loads are primarily intended for interpretation or if they will also be used in other parts of the library. Otherwise, I don't quite understand.

[oliverwolfe54-byte]

If possible, preserving the original feature names in the output would make the results much easier to interpret.

[Daniel15568]

This has been implemented, so closing this issue