Enhance skore.cross_validate: specify which metric is actually used behind test_score

[sylvaincom]

Problem

Part of epic #544

When using skore.cross_validate with linear regression for example, I can see the following scores in the legend: test_score and r2_score, while the test_score is actually the r2_score! This can be confusing for the user.

Suggested solution

It is ok the keep test_score and r2_score, but in any case, tell the user which score is actually used by default by scikit-learn for the test_score (it depends on each estimator).

[augustebaum]

So would we show this information in the plot? Or does it also have to be in the output of cross_validate itself (given the constraint that we should try to stick to scikit-learn's output format)?

[augustebaum]

How exactly should we give this information? In the title of the plot?

[sylvaincom]

Hmm subtitle of the plot?

[sylvaincom]

We need to address this, mentioned today in #1047

[sylvaincom]

In the case of the regression task, let's just display R2 (⭐️) and RMSE where the star indicates that it's scikit-learn's default

WDYT? Cc @augustebaum @MarieS-WiMLDS

[MarieS-WiMLDS]

I like the star idea because it's pretty, but is it technically possible to have an emoji in the legend? Plus, it is explicit? The other less pretty option but explicit and easy would be to print sth like R2 (sklearn default). You can do tests with Auguste and see how it seems :) !

[sylvaincom]

(We could use unicode emoji / characters that should work in matplotlib, etc, as done in #997)

After discussing with @glemaitre, this issue is on hold, waiting on #997 and #1014 as they will change the way a cross-validation report is displayed