Remove update prediction cache from predictors. (#5312)

Move this function into gbtree, and uses only updater for doing so. As now the predictor knows exactly how many trees to predict, there's no need for it to update the prediction cache.

src/common/hist_util.h

@@ -199,7 +199,7 @@ class CutsBuilder {
   }
 
   void AddCutPoint(WQSketch::SummaryContainer const& summary, int max_bin) {
-    int required_cuts = std::min(static_cast<int>(summary.size), max_bin);
+    size_t required_cuts = std::min(summary.size, static_cast<size_t>(max_bin));
     for (size_t i = 1; i < required_cuts; ++i) {
       bst_float cpt = summary.data[i].value;
       if (i == 1 || cpt > p_cuts_->cut_values_.back()) {

src/common/transform.h

@@ -181,8 +181,7 @@ class Transform {
    * \param func    A callable object, accepting a size_t thread index,
    *                  followed by a set of Span classes.
    * \param range   Range object specifying parallel threads index range.
-   * \param devices GPUSet specifying GPUs to use, when compiling for CPU,
-   *                  this should be GPUSet::Empty().
+   * \param device  Specify GPU to use.
    * \param shard Whether Shard for HostDeviceVector is needed.
    */
   template <typename Functor>