cuda discriptor added

[Leedk3]

Dear Koide3,

First and foremost, I would like to express my gratitude for sharing the fast registration algorithm. It has been an inspiring project to work on.

I wanted to inform you that I've made some significant enhancements to the algorithm by implementing CUDA-based kernel descriptors for the nearest points search. This addition has considerably improved the algorithm's performance and efficiency. (With my custom dataset, Laplasian kernel was best.)

Thank you once again for providing the foundation for this work, and I look forward to contributing to the project's progress.

Best regards, Daegyu Lee

[koide3]

Hi @Leedk3 ,

Thanks a lot for your PR. It looks a big improvement, and I would live to take it. Before taking a deeper look at the updated code, I kindly request you to update/provide the following points.

• Provide a brief explanation of the implemented descriptors and how it improves the scan matching performance
• Keep the default value of BUILD_VGICP_CUDA to OFF so that it passes the CI
• Replace comments in Korean in covariance_estimation_polynomial.cu with English ones

Thanks again for your willingness to contribute to this package :+1: