[CI] Tests (net10.0) - ModelFamily - Regression: RBF scaling-equivariance

[ooples]

Tracked from audit 2026-05-20 against PR #1404 CI run 26169970681.

This shard was never tracked by the original PR-1290 epic (#1315) but has been failing across recent PRs.

Current failing tests

• AiDotNet.Tests.ModelFamilyTests.Regression.RadialBasisFunctionRegressionTests.ScalingEquivariance_ScalingTargets_ScalesPredictions

Pattern

Single failing test. The ScalingEquivariance_ScalingTargets_ScalesPredictions invariant asserts that if you scale targets by a constant k, the trained model's predictions scale by the same k. This is a mathematical property of linear-ish regressors. Failure usually means:

1. The model has a hidden bias term not being scaled correctly
2. The regularization term is target-dependent
3. The model fits a non-linear transform that breaks equivariance

Need to trace through RadialBasisFunctionRegression.Train() to find where the scale gets dropped.