Re-implement ROC-AUC. (#6747)

* Re-implement ROC-AUC.

* Binary
* MultiClass
* LTR
* Add documents.

This PR resolves a few issues:
  - Define a value when the dataset is invalid, which can happen if there's an
  empty dataset, or when the dataset contains only positive or negative values.
  - Define ROC-AUC for multi-class classification.
  - Define weighted average value for distributed setting.
  - A correct implementation for learning to rank task.  Previous
  implementation is just binary classification with averaging across groups,
  which doesn't measure ordered learning to rank.

src/objective/rank_obj.cu

@@ -293,7 +293,7 @@ class NDCGLambdaWeightComputer
                                              group_segments)),
                             thrust::make_discard_iterator(),  // We don't care for the group indices
                             dgroup_dcg_.begin());  // Sum of the item's DCG values in the group
-    CHECK(static_cast<unsigned>(end_range.second - dgroup_dcg_.begin()) == dgroup_dcg_.size());
+    CHECK_EQ(static_cast<unsigned>(end_range.second - dgroup_dcg_.begin()), dgroup_dcg_.size());
   }
 
   inline const common::Span<const float> GetGroupDcgsSpan() const {