[ENH] add XGBoost Experiment integration

This issue discusses if we should add a thin wrapper for the XGBoost integration. XGBoost already works with the sklearn-integration, because the estimators are so similar. Does it still make sense to add a class for an XGBoost integration to show users clearly, that this package is supported in Hyperactive, or show we just document in the sklearn integration, that XGBoost is supported?