Steps to integrate third-party feature detector

[ccaven]

Hello! I appreciate your work in supporting open-source visual SLAM.
I am currently working on a (non-CUDA) GPU implementation of the ORB feature extractor, and would love to connect it to stella_vslam_ros. Here's what I think my TODO list is:

1. Add a new function in stella_vslam called system::feed_mono_keypoints / system::feed_stereo_keypoints which skips the ORB extraction and only runs undistort_keypoints then convert_keypoints_to_bearings and assign_keypoints_to_grid, etc.
2. Add a new subscription in stella_vslam_ros called camera/keypoints that then calls system::feed_mono_keypoints
3. In my own code, publish to camera/keypoints
   You know the stella_vslam codebase better than I do, so if I'm missing anything / there is something I should know, please let me know. Thanks!

[ymd-stella]

Looks good to me. One thing to point out is that the type that represents keypoints is not in the ROS/ROS2 common message types. You must create a custom message.