Improve df.corr: remove numpy; add method='spearman'; add labels; make consistent with pl.corr

[Julian-J-S]

Description

Correlation analysis is a very frequently used data-science method.

The current implementation could be improved as recommended by a few issues:

• pl.DataFrame.corr(method = "spearman") #16864
• Add the option to use Spearman's rank correlation in polars.DataFrame.corr #14457
• pl.DataFrame.corr() should offer an argument "include_column_names_as_column". #16832

Make consistent with pl.corr: Remove numpy dependency & add method=spearman

• df.corr link
  • based on numpy (corrcoef) according to the documentation
  • only has method pearson
• pl.corr link
  • native rust
  • has pearson and spearman

Add parameter to add row labels

• for correlations when data has lots of rows it would be very handy to optionally add a row label that links between the columns of the correlation

[kevuoft]

Can we also have kendall correlation to match Pandas.DataFrame.corr()?

[fraruscio]

Can we also have kendall correlation to match Pandas.DataFrame.corr()?

I also think kendall would be more useful than spearman for many practical cases, and it would be really helpful to have a native implementation for large dataframes.