Fix build on CUDA 12.6 and Pytorch >= 2.1

[keisukefukuda]

This PR is a re-submission of #22

The issue of __thrust_exec_check_disable__ macro

On my environment with CUDA 12.6, build is still broken after the change e1ea8bd .

Although I'm not sure about when __thrust_exec_check_disable__ was removed, but
I checked the NVIDIA's official Docker image and confirmed that it was already removed at 12.6.

$ docker run -it nvidia/cuda:12.6.3-cudnn-devel-ubuntu20.04 -- bash -c "grep '__thrust_exec_check_disable__' -r /usr/local/cuda/*/*/include"

==========
== CUDA ==
==========

CUDA Version 12.6.3

Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.

This container image and its contents are governed by the NVIDIA Deep Learning Container License.
By pulling and using the container, you accept the terms and conditions of this license:
https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license

A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.

__thrust_exec_check_disable__ is not found.

$ docker run -it nvidia/cuda:12.6.3-cudnn-devel-ubuntu20.04 -- bash -c "grep 'define _CCCL_EXEC_CHECK_DISABLE' -r /usr/local/cuda/*/*/include"

(snip)

/usr/local/cuda/targets/x86_64-linux/include/cuda/std/__cccl/execution_space.h:#      define _CCCL_EXEC_CHECK_DISABLE __pragma("nv_exec_check_disable")
/usr/local/cuda/targets/x86_64-linux/include/cuda/std/__cccl/execution_space.h:#      define _CCCL_EXEC_CHECK_DISABLE _Pragma("nv_exec_check_disable")
/usr/local/cuda/targets/x86_64-linux/include/cuda/std/__cccl/execution_space.h:#    define _CCCL_EXEC_CHECK_DISABLE

_CCCL_EXEC_CHECK_DISABLE is found.

So, it would be more robust to check the existence of _CCCL_EXEC_CHECK_DISABLE and __thrust_exec_check_disable__ with ifdef .

C++17

In addition, recent Pytorch versions require C++17 and I think it's safe to migrate the entire dietgpu project to C++17.