feat(metrics): add a classification metrics module

[IndexSeek]

I wanted to start making some progress on supporting a new ibis_ml.metrics
module might look like with the first pass to support classification metrics.

In this initial pass, the functions accept the y_pred and y_true as Ibis table
expression columns.

A few thoughts

• Should we go ahead and try this with regression metrics? Or would it be okay
  to start with classification only as an "experimental" type of op.
• Do we need a "metrics" folder that contains "regression" and "classification" files?
• Is there a more effective way to test this than to compare it to scikit-learns metrics?

What could go wrong?

I anticipate the inputs are coming from the same table expression, although testing
this behavior (or at least documenting) with what happens with input from two different
table expressions would be necessary.

Resolves (partially) #174

Edit: As far as testing goes, I wonder if more precision might be necessary.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.49%. Comparing base (7d2d021) to head (2b57d2b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #176      +/-   ##
==========================================
+ Coverage   90.35%   90.49%   +0.14%     
==========================================
  Files          29       31       +2     
  Lines        2083     2114      +31     
==========================================
+ Hits         1882     1913      +31     
  Misses        201      201              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[deepyaman]

Sorry, I didn't see this earlier; will try to take a look soon!

[deepyaman]

One functional request to not force compute; rest are quick nits!

[IndexSeek]

Thanks for your review and the suggestions. I think these were great and have added them all now.

I'm not sure what's going on with the build-docs check here, it seems the last time it succeeded it was using scikit-learn==1.5.2 and now it's using scikit-learn==1.6.0.

[deepyaman]

I'm not sure what's going on with the build-docs check here, it seems the last time it succeeded it was using scikit-learn==1.5.2 and now it's using scikit-learn==1.6.0.

Seems __sklearn_tags__ is new from scikit-learn/scikit-learn#29677; I'll have to look into it. Feel free to pin scikit-learn <1.6.0 for now in https://github.com/ibis-project/ibis-ml/blob/main/pyproject.toml#L23.

[IndexSeek]

It's pinned now at <1.6.0 for the "doc" dependencies and the check is passing! ✅

[deepyaman]

LGTM!