The lanelet2_map_filter_node calculates slowly. What can be done to make this node calculate faster?

[0116lhc]

I've noticed that the calculation speed of this node for filtering point clouds is a bit slow, especially when testing the filtering of the original point clouds（/sensing/lidar/concatenated/pointcloud）. I took a look at the source code and found that it's about determining whether the downsampled point clouds are within the regions of interest. When we have a lot of boxes drawn for the vector_map, the calculation speed will slow down and affect the subsequent nodes.
I've come up with two ideas:
Is it possible to combine the regions of interest or directly merge all the boxes of the entire vector_map into one box? In this way, the point cloud filtering would only need to be calculated once, but it would take extra time to merge all the vector_map boxes into one.
The voxel_size_x and voxel_size_y set in the example routine are both 0.04. Can these be set to 1.0 or even larger values to reduce the amount of calculation? Is this feasible?
May I ask which of these two methods is better? I'm going to try the second one first. I need your help to try the first one. Thank you.