onnxruntime [Build] CCCL API migration issue.

Describe the issue

We would like to admit that following deprecation required to be completed. https://nvidia.github.io/cccl/cccl/3.0_migration_guide.html

We eventually start seeing issue during the compilation process against product that are in development for NVIDIA.

Urgency

It need to be patched before June.

Target platform

Linux

Build script

build [DEBUG] - Command line arguments:
  --build_dir /workspace/onnxruntime/build/Linux --config Release --skip_submodule_sync --parallel --build_shared_lib --compile_no_warning_as_error --build_dir /workspace/build --cmake_extra_defines 'CMAKE_CUDA_ARCHITECTURES='"'"'75;80;86;89;90;100;120'"'"'' --cmake_extra_defines CMAKE_POLICY_VERSION_MINIMUM=3.5 --update --build --use_cuda --cuda_home /usr/local/cuda --cudnn_home /usr --use_tensorrt --use_tensorrt_builtin_parser --tensorrt_home /usr/src/tensorrt --allow_running_as_root --use_openvino CPU

Error / output

[ 80%] Building CUDA object CMakeFiles/onnxruntime_providers_cuda.dir/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/cudnn_fmha/cudnn_flash_attention.cu.o
/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(98): error: namespace "cub" has no member "Max"
    const auto max = BlockReduce(tmp_storage).Reduce(thread_data_max, cub::Max());
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(117): error: namespace "cub" has no member "Sum"
    const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_sum, cub::Sum());
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(174): error: namespace "cub" has no member "Max"
    const auto max = BlockReduce(tmp_storage).Reduce(input_data, cub::Max(), end);
                                                                      ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(187): error: namespace "cub" has no member "Sum"
    const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_exp, cub::Sum(), end);
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(243): error: namespace "cub" has no member "Max"
    const auto max = BlockReduce(tmp_storage).Reduce(thread_data_max, cub::Max(), end);
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(257): error: namespace "cub" has no member "Sum"
    const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_exp, cub::Sum(), end);
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(346): error: namespace "cub" has no member "Max"
    const float max = BlockReduce(tmp_storage).Reduce(max_thread_data, cub::Max(), total_sequence_length);
                                                                            ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(360): error: namespace "cub" has no member "Sum"
    const auto sum = BlockReduce(tmp_storage).Reduce(sum_thread_data_exp, cub::Sum(), TPB);
                                                                               ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(444): error: namespace "cub" has no member "Max"
    const float max = BlockReduce(tmp_storage).Reduce(thread_data, cub::Max(), total_sequence_length);
                                                                        ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(453): error: namespace "cub" has no member "Sum"
    const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_exp, cub::Sum(), total_sequence_length);
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(599): error: namespace "cub" has no member "Max"
    const auto max = BlockReduce(tmp_storage).Reduce(thread_data_max, cub::Max(), end);
                                                                           ^

/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu(612): error: namespace "cub" has no member "Sum"
    const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_exp, cub::Sum(), end);
                                                                           ^

12 errors detected in the compilation of "/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu".
gmake[2]: *** [CMakeFiles/onnxruntime_providers_cuda.dir/build.make:5978: CMakeFiles/onnxruntime_providers_cuda.dir/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/attention_softmax.cu.o] Error 2
gmake[2]: *** Waiting for unfinished jobs....
/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/bert_padding.cu(386): error: namespace "cub" has no member "Max"
    int last_leading_position = BlockReduce(temp_storage).Reduce(biggest_position, cub::Max(), blockDim.x);
                                                                                        ^

1 error detected in the compilation of "/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/bert_padding.cu".
gmake[2]: *** [CMakeFiles/onnxruntime_providers_cuda.dir/build.make:6023: CMakeFiles/onnxruntime_providers_cuda.dir/workspace/onnxruntime/onnxruntime/contrib_ops/cuda/bert/bert_padding.cu.o] Error 2