Fix shuffle comm operators with no default constructors

[caugonnet]

Operators such as key_group_id_less_t are templated by functors which we instantiate using lambda functions. These lambda functions may have capture lists so that there is no default constructor, and we cannot call the default constructor of this operator.

This solves the following errors :

/home/caugonnet/git/caugonnet_cugraph/cpp/include/cugraph/utilities/shuffle_comm.cuh(172): error: the default constructor of "lambda [](auto)->auto" cannot be referenced -- it is a deleted function
    ValueToGroupIdOp value_to_group_id_op{};
                                         ^
          detected during:
            instantiation of class "cugraph::detail::value_group_id_greater_equal_t<value_type, ValueToGroupIdOp> [with value_type=int, ValueToGroupIdOp=lambda [](auto)->auto]" at line 626 of /home/caugonnet/git/caugonnet_cugraph/cpp/build/_deps/cccl-src/lib/cmake/libcudacxx/../../../libcudacxx/include/cuda/std/detail/libcxx/include/tuple
            instantiation of class "cuda::std::__4::__tuple_constraints<_Tp...> [with _Tp=<cugraph::detail::value_group_id_greater_equal_t<int, lambda [](auto)->auto>>]" at line 629 of /home/caugonnet/git/caugonnet_cugraph/cpp/build/_deps/cccl-src/lib/cmake/libcudacxx/../../../libcudacxx/include/cuda/std/detail/libcxx/include/tuple
            instantiation of class "cuda::std::__4::__tuple_constraints<_Tp...> [with _Tp=<cugraph::detail::value_group_id_greater_equal_t<int, lambda [](auto)->auto>>]" at line 157 of /home/caugonnet/git/caugonnet_cugraph/cpp/build/_deps/cccl-src/lib/cmake/libcudacxx/../../../libcudacxx/include/cuda/std/__type_traits/is_constructible.h
            instantiation of "const __nv_bool cuda::std::__4::is_constructible_v [with _Tp=cuda::std::__4::tuple<cugraph::detail::value_group_id_greater_equal_t<int, lambda [](auto)->auto>>, _Args=<>]" at line 51 of /home/caugonnet/git/caugonnet_cugraph/cpp/build/_deps/cccl-src/lib/cmake/libcudacxx/../../../libcudacxx/include/cuda/std/__functional/not_fn.h
            instantiation of class "cuda::std::__4::__not_fn_t<_Fn> [with _Fn=cugraph::detail::value_group_id_greater_equal_t<int, lambda [](auto)->auto>]" at line 56 of /home/caugonnet/git/caugonnet_cugraph/cpp/build/_deps/cccl-src/lib/cmake/libcudacxx/../../../libcudacxx/include/cuda/std/__functional/not_fn.h
            [ 3 instantiation contexts not shown ]
            instantiation of "ValueIterator cugraph::detail::mem_frugal_partition(ValueIterator, ValueIterator, ValueToGroupIdOp, int, rmm::cuda_stream_view) [with ValueIterator=int32_t *, ValueToGroupIdOp=lambda [](auto)->auto]" at line 615
            instantiation of "void cugraph::detail::mem_frugal_groupby(ValueIterator, ValueIterator, ValueToGroupIdOp, int, size_t, rmm::cuda_stream_view) [with ValueIterator=int32_t *, ValueToGroupIdOp=lambda [](auto)->auto]" at line 775
            instantiation of "rmm::device_uvector<size_t> cugraph::groupby_and_count(ValueIterator, ValueIterator, ValueToGroupIdOp, int, size_t, rmm::cuda_stream_view) [with ValueIterator=int32_t *, ValueToGroupIdOp=lambda [](auto)->auto]" at line 42 of /home/caugonnet/git/caugonnet_cugraph/cpp/src/utilities/shuffle_vertices.cuh
            instantiation of "rmm::device_uvector<vertex_t> cugraph::<unnamed>::shuffle_vertices_by_gpu_id_impl(const raft::handle_t &, rmm::device_uvector<vertex_t> &&, func_t) [with vertex_t=int32_t, func_t=cugraph::detail::compute_gpu_id_from_int_vertex_t<int32_t>]" at line 158 of /home/caugonnet/git/caugonnet_cugraph/cpp/src/utilities/shuffle_vertices.cuh
            instantiation of "rmm::device_uvector<vertex_t> cugraph::detail::shuffle_int_vertices_to_local_gpu_by_vertex_partitioning(const raft::handle_t &, rmm::device_uvector<vertex_t> &&, const std::vector<vertex_t, std::allocator<vertex_t>> &) [with vertex_t=int32_t]" at line 31 of /home/caugonnet/git/caugonnet_cugraph/cpp/src/utilities/shuffle_vertices_mg_v32_integral.cu

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[ChuckHastings]

/ok to test

[seunghwak]

Do we still need this once #4888 gets merged?

[caugonnet]

I guess with #4888 the constructors are implicit ... I haven't check if all classes have been modified

[ChuckHastings]

Can we close this in favor of #4888

[ChuckHastings]

With #4888 merged, can we close this or is there something remaining to do?