[ML] Add unit test checks for new hyperparameters in supervised learning

[valeriy42]

For the tests with all hyperparameters set (e.g. m_RegularizationOverride.*), we need to check that those would set the actual hyperparameter values (e.g. m_Regularization.*). I just noticed that setting softTreeDepthLimit and softTreeDepthTolerance was missing and the unit test testRunBoostedTreeRegressionTrainingWithParams didn't fail.

[tveasey]

Good catch!