[R-package] a few fixes for R (#1485)

* [R] fix #1465

* [R] add sanity check to fix #1434

* [R] some clean-ups for custom obj&eval; require maximize only for early stopping

R-package/R/callbacks.R

@@ -458,6 +458,7 @@ cb.save.model <- function(save_period = 0, save_name = "xgboost.model") {
 #' \code{basket},
 #' \code{data},
 #' \code{end_iteration},
+#' \code{params},
 #' \code{num_parallel_tree},
 #' \code{num_class}.
 #' 
@@ -491,6 +492,9 @@ cb.cv.predict <- function(save_models = FALSE) {
 
     ntreelimit <- NVL(env$basket$best_ntreelimit, 
                       env$end_iteration * env$num_parallel_tree)
+    if (NVL(env$params[['booster']], '') == 'gblinear') {
+      ntreelimit <- 0 # must be 0 for gblinear
+    }
     for (fd in env$bst_folds) {
       pr <- predict(fd$bst, fd$watchlist[[2]], ntreelimit = ntreelimit, reshape = TRUE)
       if (is.matrix(pred)) {