[FEA] Look at using new copy_if/count_if APIs

[revans2]

Is your feature request related to a problem? Please describe.
rapidsai/cudf#20861 is looking to replace copy_if and count_if in thrust with something better. We have some calls to these APIs in spark-rapids-jni. When that PR goes in we should look at migrating to cudf::detail::copy_if and cudf::detail::count_if

[revans2]

Looks like these are changes we need to make to get the code to compile as copy_if_safe was removed.

FAILED: CMakeFiles/spark_rapids_jni.dir/src/from_json_to_raw_map.cu.o 
/usr/local/cuda/bin/nvcc -forward-unknown-to-host-compiler -DCCCL_DISABLE_PDL -DCUB_DISABLE_NAMESPACE_MAGIC -DCUB_IGNORE_NAMESPACE_MAGIC_ERROR -DCUDA_API_PER_THREAD_DEFAULT_STREAM -DCUDF_USE_PER_THREAD_DEFAULT_STREAM -DFMT_HEADER_ONLY=1 -DLIBCUDACXX_ENABLE_EXPERIMENTAL_MEMORY_RESOURCE -DSPDLOG_FMT_EXTERNAL -DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_CUDA -DTHRUST_DISABLE_ABI_NAMESPACE -DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -DTHRUST_IGNORE_ABI_NAMESPACE_ERROR -Dspark_rapids_jni_EXPORTS -I/__w/cudf/cudf/src/main/cpp/src -I/usr/lib/jvm/java/include -I/usr/lib/jvm/java/include/linux -I/__w/cudf/cudf/thirdparty/cudf/java/src/main/native/include -I/__w/cudf/cudf/thirdparty/cudf/java/src/main/native/src -I/__w/cudf/cudf/target/libcudfjni/_deps/arrow-src/cpp/src -I/__w/cudf/cudf/target/libcudfjni/_deps/thrift-build -I/__w/cudf/cudf/target/libcudfjni/_deps/thrift-src/lib/cpp/src -I/__w/cudf/cudf/target/libcudf-install/include/rapids -isystem /usr/local/cuda/targets/x86_64-linux/include -isystem /__w/cudf/cudf/target/libcudf-install/include -O3 -DNDEBUG -std=c++20 "--generate-code=arch=compute_90,code=[compute_90,sm_90]" -Xcompiler=-fPIC --expt-extended-lambda --expt-relaxed-constexpr -Werror=all-warnings -Xcompiler=-Wall,-Werror,-Wno-error=deprecated-declarations -diag-suppress=1407 -Xfatbin=-compress-all -Xfatbin=--compress-level=3 -MD -MT CMakeFiles/spark_rapids_jni.dir/src/from_json_to_raw_map.cu.o -MF CMakeFiles/spark_rapids_jni.dir/src/from_json_to_raw_map.cu.o.d -x cu -c /__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu -o CMakeFiles/spark_rapids_jni.dir/src/from_json_to_raw_map.cu.o
/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(565): error: no instance of function template "cudf::detail::copy_if_safe" matches the argument list
            argument types are: (cudf::size_type *, cudf::size_type *, cudf::size_type *, cuda::__4::__detail::__return_type_wrapper<bool, lambda [](auto)->auto>, rmm::cuda_stream_view)
    auto const copy_end = cudf::detail::copy_if_safe(
                          ^
/__w/cudf/cudf/target/libcudf-install/include/cudf/detail/utilities/algorithm.cuh(41): note #3322-D: number of parameters of function template "cudf::detail::copy_if_safe" does not match the call
  OutputIterator copy_if_safe(InputIterator first,
                 ^

/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(571): error: no instance of function template "cuda::std::__4::distance" matches the argument list
            argument types are: (cudf::size_type *, const <error-type>)
   cuda::std::distance(list_offsets.begin(), copy_end) == static_cast<int64_t>(n_lists)
   ^
/__w/cudf/cudf/target/libcudf-install/include/rapids/cuda/std/__iterator/distance.h(40): note #3327-D: candidate function template "cuda::std::__4::distance" failed deduction
  distance(_InputIter __first, _InputIter __last)
  ^

/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(639): error: no instance of function template "cudf::detail::copy_if_safe" matches the argument list
            argument types are: (const thrust::counting_iterator<cudf::size_type, thrust::use_default, thrust::use_default, thrust::use_default, thrust::detail::unit_stride>, thrust::counting_iterator<cudf::size_type, thrust::use_default, thrust::use_default, thrust::use_default, thrust::detail::unit_stride>, cudf::size_type *, spark_rapids_jni::<unnamed>::is_invalid_struct_begin, rmm::cuda_stream_view)
      cudf::detail::copy_if_safe(node_id_it,
      ^
/__w/cudf/cudf/target/libcudf-install/include/cudf/detail/utilities/algorithm.cuh(41): note #3322-D: number of parameters of function template "cudf::detail::copy_if_safe" does not match the call
  OutputIterator copy_if_safe(InputIterator first,
                 ^

/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(644): error: no instance of function template "cuda::std::__4::distance" matches the argument list
            argument types are: (cudf::size_type *, const <error-type>)
    auto const num_invalid = cuda::std::distance(invalid_indices.begin(), invalid_copy_end);
                             ^
/__w/cudf/cudf/target/libcudf-install/include/rapids/cuda/std/__iterator/distance.h(40): note #3327-D: candidate function template "cuda::std::__4::distance" failed deduction
  distance(_InputIter __first, _InputIter __last)
  ^

/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(659): error: no instance of function template "cudf::detail::copy_if_safe" matches the argument list
            argument types are: (const thrust::counting_iterator<cudf::size_type, thrust::use_default, thrust::use_default, thrust::use_default, thrust::detail::unit_stride>, thrust::counting_iterator<cudf::size_type, thrust::use_default, thrust::use_default, thrust::use_default, thrust::detail::unit_stride>, cudf::size_type *, spark_rapids_jni::<unnamed>::is_line_begin, rmm::cuda_stream_view)
        cudf::detail::copy_if_safe(node_id_it,
        ^
/__w/cudf/cudf/target/libcudf-install/include/cudf/detail/utilities/algorithm.cuh(41): note #3322-D: number of parameters of function template "cudf::detail::copy_if_safe" does not match the call
  OutputIterator copy_if_safe(InputIterator first,
                 ^

/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(665): error: no instance of function template "cuda::std::__4::distance" matches the argument list
            argument types are: (cudf::size_type *, const <error-type>)
        cuda::std::distance(line_begin_indices.begin(), line_begin_copy_end);
        ^
/__w/cudf/cudf/target/libcudf-install/include/rapids/cuda/std/__iterator/distance.h(40): note #3327-D: candidate function template "cuda::std::__4::distance" failed deduction
  distance(_InputIter __first, _InputIter __last)
  ^

/__w/cudf/cudf/src/main/cpp/src/from_json_to_raw_map.cu(675): error: no instance of overloaded function "thrust::for_each" matches the argument list
            argument types are: (rmm::exec_policy_nosync, cudf::size_type *, <error-type>, lambda [](auto)->auto)
      thrust::for_each(rmm::exec_policy_nosync(stream),
      ^
/__w/cudf/cudf/target/libcudf-install/include/rapids/thrust/detail/for_each.inl(51): note #3322-D: number of parameters of function template "thrust::for_each(InputIterator, InputIterator, UnaryFunction)" does not match the call
  InputIterator for_each(InputIterator first, InputIterator last, UnaryFunction f)
                ^
/__w/cudf/cudf/target/libcudf-install/include/rapids/thrust/detail/for_each.inl(38): note #3327-D: candidate function template "thrust::for_each(const thrust::detail::execution_policy_base<DerivedPolicy> &, InputIterator, InputIterator, UnaryFunction)" failed deduction
  __attribute__((host)) __attribute__((device)) InputIterator for_each(

[revans2]

Okay I just spoke with someone from CUDF and the recommendation is

Would recommend replacing both thrust::copy_if and cudf::detail::copy_if_safe (non-stencil overloads) with cudf::detail::copy_if

But we need to do this once the cudf PR is merged in.

[pmattione-nvidia]

I've convinced them to break up the PR into two parts: one that introduces the new calls and a future one to remove copy_if_safe once we switch away from it. so our builds shouldn't get broken