Use `UpdateAllowUnknown' for non-model related parameter. (#4961)

* Use `UpdateAllowUnknown' for non-model related parameter.

Model parameter can not pack an additional boolean value due to binary IO
format.  This commit deals only with non-model related parameter configuration.

* Add tidy command line arg for use-dmlc-gtest.

src/metric/elementwise_metric.cu

@@ -327,7 +327,6 @@ struct EvalEWiseBase : public Metric {
     CHECK_EQ(preds.Size(), info.labels_.Size())
         << "label and prediction size not match, "
         << "hint: use merror or mlogloss for multi-class classification";
-    const auto ndata = static_cast<omp_ulong>(info.labels_.Size());
     int device = tparam_->gpu_id;
 
     auto result =

src/metric/metric_common.h

@@ -5,7 +5,6 @@
 #ifndef XGBOOST_METRIC_METRIC_COMMON_H_
 #define XGBOOST_METRIC_METRIC_COMMON_H_
 
-#include <dmlc/parameter.h>
 #include "../common/common.h"
 
 namespace xgboost {

src/metric/multiclass_metric.cu

@@ -172,7 +172,6 @@ struct EvalMClassBase : public Metric {
     CHECK_GE(nclass, 1U)
         << "mlogloss and merror are only used for multi-class classification,"
         << " use logloss for binary classification";
-    const auto ndata = static_cast<bst_omp_uint>(info.labels_.Size());
 
     int device = tparam_->gpu_id;
     auto result = reducer_.Reduce(*tparam_, device, nclass, info.weights_, info.labels_, preds);