Currently when CUDA_ARCH_BIN isn't specified CMake attempts to build .cu files for the all architechtures it knows about, generating a list of architectures supported by the current CUDA Toolkit (see #17432). The idea being that if a user doesn't specify the arch then to be on the safe side it should generate for all architechtures for maximum compatibility. Additionally if a user just wants to specify a PTX architechture they have to pass CUDA_ARCH_BIN= (incurring the overhead of calling ocv_filter_available_architecture on all know architectures) in addition to CUDA_ARCH_PTX=<TAERGET ARCH> to CMake to avoid generating binary code for all supported architectures as well.

This PR proposes that:

1. When CUDA_ARCH_BIN and CUDA_ARCH_PTX are not specified, in addition to binary code for all compute capabilities, PTX code for the highest supported architecture is also generated. This would ensure that the CUDA device code will also run on newer GPU's than the installed CUDA Toolkit, increasing the level of compatibility futher.
2. When CUDA_ARCH_PTX is passed in isolation only PTX code for the desired CC is generated.

@tomoaki0705

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
• The PR is proposed to the proper branch
• There is a reference to the original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake