GPU Metal: also run the border-range sorts on the host

The border-range sorts of the track merging between sectors, 360 per
time frame, took 312 ms per event on a 30-collision Pb-Pb time frame on
a single GPU thread. On the host they take about 6 ms, plus 52 ms of
waiting for the streams.

Author: ktf

GPU/GPUTracking/Base/metal/GPUReconstructionMetalKernelsSpecialize.inc

@@ -83,3 +83,13 @@ inline void GPUReconstructionMetal::runKernelBackend<GPUTPCGMO2Output, GPUTPCGMO
 {
   SortOnHost(_xyz, mProcessorsShadow->tpcMerger.TrackSortO2(), [&] { return HostPointer(mProcessorsShadow->tpcMerger.Memory())->nO2Tracks; }, internal::GPUTPCGMO2OutputSort_comp());
 }
+
+template <>
+inline void GPUReconstructionMetal::runKernelBackend<GPUTPCGMMergerMergeBorders, 3>(const krnlSetupTime& _xyz, gputpcgmmergertypes::GPUTPCGMBorderRange* const& range, int32_t const& N, int32_t const& cmpMax)
+{
+  if (cmpMax) {
+    SortOnHost(_xyz, range, [&] { return N; }, internal::MergeBorderTracks_compMax());
+  } else {
+    SortOnHost(_xyz, range, [&] { return N; }, internal::MergeBorderTracks_compMin());
+  }
+}