Revamp the rabit implementation. (#10112)

This PR replaces the original RABIT implementation with a new one, which has already been partially merged into XGBoost. The new one features:
- Federated learning for both CPU and GPU.
- NCCL.
- More data types.
- A unified interface for all the underlying implementations.
- Improved timeout handling for both tracker and workers.
- Exhausted tests with metrics (fixed a couple of bugs along the way).
- A reusable tracker for Python and JVM packages.

tests/cpp/data/test_metainfo.cc

@@ -3,15 +3,16 @@
  */
 #include "test_metainfo.h"
 
-#include <gmock/gmock.h>
 #include <dmlc/io.h>
+#include <gmock/gmock.h>
 #include <xgboost/data.h>
 
 #include <memory>
 #include <string>
 
-#include "../filesystem.h"  // dmlc::TemporaryDirectory
-#include "../helpers.h"     // for GMockTHrow
+#include "../collective/test_worker.h"  // for TestDistributedGlobal
+#include "../filesystem.h"              // dmlc::TemporaryDirectory
+#include "../helpers.h"                 // for GMockTHrow
 #include "xgboost/base.h"
 
 namespace xgboost {
@@ -118,8 +119,8 @@ void VerifyGetSetFeatureColumnSplit() {
 }  // anonymous namespace
 
 TEST(MetaInfo, GetSetFeatureColumnSplit) {
-  auto constexpr kWorldSize{3};
-  RunWithInMemoryCommunicator(kWorldSize, VerifyGetSetFeatureColumnSplit);
+  auto constexpr kWorkers{3};
+  collective::TestDistributedGlobal(kWorkers, VerifyGetSetFeatureColumnSplit);
 }
 
 TEST(MetaInfo, SaveLoadBinary) {

tests/cpp/data/test_simple_dmatrix.cc

@@ -9,6 +9,7 @@
 
 #include "../../../src/data/adapter.h"         // ArrayAdapter
 #include "../../../src/data/simple_dmatrix.h"  // SimpleDMatrix
+#include "../collective/test_worker.h"         // for TestDistributedGlobal
 #include "../filesystem.h"                     // dmlc::TemporaryDirectory
 #include "../helpers.h"                        // RandomDataGenerator,CreateSimpleTestData
 #include "xgboost/base.h"
@@ -444,5 +445,5 @@ void VerifyColumnSplit() {
 
 TEST(SimpleDMatrix, ColumnSplit) {
   auto constexpr kWorldSize{3};
-  RunWithInMemoryCommunicator(kWorldSize, VerifyColumnSplit);
+  collective::TestDistributedGlobal(kWorldSize, VerifyColumnSplit);
 }