ranger and xgboost don't respect R's seed

[jeremyrcoyle]

Instead, each takes a separate seed argument. Currently, users can specify these manually as parameters:

lrnr_ranger <- make_learner(Lrnr_ranger, min.node.size = 30, num.trees=500, max.depth=2, seed=1234)
lrnr_xgboost <- make_learner(Lrnr_xgboost, minobspernode = 30, ntrees=500, max_depth=2, subsample=1, seed=1234)

However, we should manually generate and set seeds for these learners as a function of the current R seed to make reproducibility easier. Thanks to @yqzhong7 for reporting issues with reproducibility and these learners