Add argument to `surv.xgboost` to differentiate between Cox and AFT objectives

[bblodfon]

• Example => https://github.com/mlr-org/mlr3torch/blob/main/R/LearnerTorchMLP.R
• This will allow us to add distr prediction using Breslow estimator to Cox

[sebffischer]

You can still add two different versions to the dictionary. mlr_learners$add("id1", function() LearnerSurvXGBoost$new(type = 1))

mlr_learners$add("id2", function() LearnerSurvXGBoost$new(type = 2))

otherwise the Dictionary -> data.table conversion might show slightly imprecise information on the learner regarding predict types :)