Update paramtest logic and update various learner param sets

[pat-s]

Paramtest logic

As discussed with @be-marc

• Merge train and control functions into one test
• Use tag = "train" and tag = "predict" to select the correct params in each test
• Use the paramtest resources from the respective branch instead of the CRAN system files (otherwise changes made in the branch are not evaluated in a PR)

Param set updates

xgboost

• disable_default_eval_metric (NEW)
• monotone_constraints (NEW)
• process_type (NEW)
• sampling_method (NEW)
• single_precision_histogram (NEW)
• seed_per_iteration (NEW)

glmnet & cv.glmnet

• standardize.response (NEW)
• type.gaussian (NEW)

svm

• epsilon (NEW)

multinom

• entropy (NEW)
• mask (NEW)
• MaxNWts (NEW)
• linout (NEW)
• size (NEW)
• skip (NEW)
• softmax (NEW)
• Wts (NEW)

ranger

• seed (also for predict)

[pat-s]

@RaphaelS1 I've also updated the surv learners - just in case you want to have a look :)

[RaphaelS1]

@RaphaelS1 I've also updated the surv learners - just in case you want to have a look :)

Thanks @pat-s , surv changes look good to me