aorsf time prediction

[bcjaeger]

See https://github.com/ropensci/aorsf/issues/49

• Updates default oobag prediction type to match aorsf v0.1.3
• Adds code for survival time prediction with aorsf, and a test.

[bblodfon]

Hi @bcjaeger, thanks for the PR! Some notes/suggestions:

1. The oobag_pred_type = mort is supported in v0.1.3 right? (I don't see it in the available options)
2. From a quick check on the learner's parameters I noticed that tree_seeds and oobag_fun are not included, any particular reason why is that?
3. Can we add a bit of documentation on the returned prediction types? Like here for example. We can add this ref for the RMST!
4. For the RMST (response) if you have a function that can do it automatically from your package then we could use it, instead of writing the code manually (both things work) - what is more preferable @sebffischer ?

[bcjaeger]

Hello! Thanks for reviewing and following up. I incorporated all the suggestions except for including an option to set oobag_fun. The reason is that I'm not sure how I should set up an input parameter that is a function.

[bblodfon]

@bcjaeger thanks for the new CRAN version! The issue with the survival time is fixed there. Please add the new parameter as well, I get this now (the rest are our problems):

Missing parameters:
- 'oobag'

[bblodfon]

paradox update 10/6, then fixing and merging this

[bblodfon]

@sebffischer only thing left here is adding the early_stopping_min_delta in lightGBM - I hesitate to do it because early stopping might have its perks (I saw a train-specific function), can I leave it to you?