Update `boostSplines` and `boostLinear`

[schalkdaniel]

Add the option, e.g., when iteration = 0 to return the untrained model. This would allow the user to add a further base learner or logger and then train the model.