Remove `switch` arg in `DropHighPSIFeatures`

[sancholp]

Closes #858.

PSI is a symmetric metric and as such the switch arg doesn't add functionality.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 98.04%. Comparing base (ead2457) to head (202ae5c). Report is 6 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #859   +/-   ##
=======================================
  Coverage   98.04%   98.04%           
=======================================
  Files         111      111           
  Lines        4643     4648    +5     
  Branches      739      740    +1     
=======================================
+ Hits         4552     4557    +5     
  Misses         56       56           
  Partials       35       35           

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

:rocket: New features to boost your workflow:

• :snowflake: Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[sancholp]

@solegalli Whenever you find time please have a look

[fkiraly]

two remarks from me (just browsing through PR):

• I think it is indeed correct, PSI is symmetric. To see this, switch the two batches and observe that the difference term and the log term both switch signs.
  • however, it might not be true for the estimator, or if cutoffs are dynamically determined (in this case the switch will not result in equal value) - have you checked this?
• removing an existing argument from an sklearn class is a breaking change. If this would be sklearn, it would have to go through a deprecation, i.e., a warning is raised that the argument will be removed, so users can adapt and their code does not break in case they use the arg.

[solegalli]

As per discussion in #858, we won't remove the argument, we'll make it clear in the docs that the PSI is symmetric, but the estimator calculates it is not.

[solegalli]

Waiting for this PR https://github.com/sancholp/feature_engine/pull/1 to merge.

[solegalli]

For some reason, this didn't work. I reverted the files to the original ones to only change the docstrings but something went wrong somewhere.

[sancholp]

Anything you need me to do?