[INFRA] Import NVIDIA/CCCL upstream as optimization reference library

CCCL (CUDA C++ Core Libraries) provides:
- CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk)
- Thrust: high-level parallel algorithms (transform_reduce, sort, scan)
- libcudacxx: CUDA C++ standard library (atomics, barriers, memory)
- cudax: experimental features (memory resources, allocators)
- Tuning policies: per-SM hardware-specific algorithm parameters

Competition optimization vectors mapped to CCCL:
- Output TPS (83% weight): warp_reduce, block_reduce, device_topk
- Input TPS (14% weight): device_scan, block_load, prefetch
- Cache TPS (3% weight): prefix caching strategy patterns
- Memory (0.9 util): pooled/cached/buddy allocators

Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only)
License: Apache-2.0

cccl_upstream/cub/cmake/CubUtilities.cmake

@@ -0,0 +1,41 @@
+# cub_configure_cuda_target(<target_name> RDC <ON|OFF>)
+#
+# Configures `target_name` with the appropriate CUDA architectures and RDC state.
+function(cub_configure_cuda_target target_name)
+  set(options)
+  set(one_value_args RDC)
+  set(multi_value_args)
+  cmake_parse_arguments(
+    cub_cuda
+    "${options}"
+    "${one_value_args}"
+    "${multi_value_args}"
+    ${ARGN}
+  )
+
+  if (cub_cuda_UNPARSED_ARGUMENTS)
+    message(
+      AUTHOR_WARNING
+      "Unrecognized arguments passed to cub_configure_cuda_target: "
+      ${cub_cuda_UNPARSED_ARGUMENTS}
+    )
+  endif()
+
+  if (NOT DEFINED cub_cuda_RDC)
+    message(AUTHOR_WARNING "RDC option required for cub_configure_cuda_target.")
+  endif()
+
+  if (cub_cuda_RDC)
+    set_target_properties(
+      ${target_name}
+      PROPERTIES #
+        CUDA_SEPARABLE_COMPILATION ON
+        POSITION_INDEPENDENT_CODE ON
+    )
+  else()
+    set_target_properties(
+      ${target_name}
+      PROPERTIES CUDA_SEPARABLE_COMPILATION OFF
+    )
+  endif()
+endfunction()