parallel calculating (multithreading with omp or tbb) for VoxelDownSampleAndTrace

Thanks for the library. It is very nice and powerful.
I used the function VoxelDownSampleAndTrace and I found it very low for big point cloud (about 10M points).
I read its implementation. You didn't use any multi threadding (like other functions for example SearchKNN or RemoveStatisticalOutliers.

Is it possible to implement the multi threadding for VoxelDownSampleAndTrace ?
I know it is difficult since you use std::unordered_map, which is not thread-safety.

Thank you.