docs: Add ADR on neg_* metrics (#1051)

Closes #1020

---------

Co-authored-by: Thomas S. <[email protected]>
Co-authored-by: Sylvain Combettes <[email protected]>

docs/design/0002-never-present-neg-metrics-from-sklearn.md

@@ -0,0 +1,27 @@
+---
+status: accepted
+date: 2025-01-06
+decision-makers: ["@augustebaum", "@sylvaincom", "@glemaitre"]
+consulted: ["@ogrisel"]
+---
+
+# Never show `neg_*` metrics from sklearn
+
+## Context and Problem Statement
+
+We show various metrics to users, many directly using sklearn.
+In sklearn, many metrics are multiplied by -1 and prefixed with `neg_`, with the purpose of making all metrics "higher-is-better". This way, optimization tools in sklearn such as `GridSearchCV` do not need to figure out which way the metric should be optimized.
+This is specific to sklearn, and there is no reason to port this design over to skore.
+
+## Decision Drivers
+
+* Our data-science-literate collaborators (@ogrisel, @glemaitre, @sylvaincom) consider the `neg_` trick should remain a solution to a sklearn-specific problem, and not be displayed in plots for the skore user.
+
+## Decision Outcome
+
+Chosen option: Never show `neg_*` metrics from sklearn in skore, only use the positive counterparts. This makes reports clearer.
+
+### Consequences
+
+* We show the most relevant information to the user.
+* We might have to take on the responsibility of maintaining the "metric is higher-is-better" pre-condition ourselves.