perf(autoware_ndt_scan_matcher): replacement of pcl::KdTreeFLANN with nanoflann

[taisa1]

Description

This PR speeds up callback_sensor_points_main() on autoware_ndt_scan_matcher without changing the logical output.

The tail latency gets about x1.65 faster with this PR merged.

In this PR, I propose to replace pcl::KdTreeFLANN with nanoflann in this node.
nanoflann is a header-only kdtree library, a fork of FLANN used in PCL.

It is provided under the BSD license. kdtree_nanoflann.hpp contains code from nanoflann, so the license is displayed.

nanoflann.hpp is a copy from nanoflann v1.6.3, and is currently placed in include directory of this node, but will need to be moved if it is used on other nodes in the future.

Measurement Condition

• Ubuntu22.04 + ROS2 Humble + Autoware Universe rosbag simulation
• not core isolated / not core flequency fixed (so multi-threaded)

Related links

• nanoflann

Private Links:

• Performance analysis of previous implementation (TIER IV Internal link)
• Description, tests, measurements of nanoflann (TIER IV Internal link)

How was this PR tested?

• I set up an environment that fixes the input to the node, and confirmed that changing to nanoflann results in an increase in speed without changing the output (details in TIER IV internal).
• I confirmed that it can be run normally with sample rosbag simulation.

Notes for reviewers

None.

Interface changes

None.

Effects on system behavior

None.

[github-actions]

Thank you for contributing to the Autoware project!

🚧 If your pull request is in progress, switch it to draft mode.

Please ensure:

• You've checked our contribution guidelines.
• Your PR follows our pull request guidelines.
• All required CI checks pass before marking the PR ready for review.