[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/thrust/testing/tuple_reduce.cu

@@ -0,0 +1,56 @@
+#include <thrust/reduce.h>
+#include <thrust/transform.h>
+
+#include <cuda/std/tuple>
+
+#include <unittest/unittest.h>
+
+using namespace unittest;
+
+struct SumTupleFunctor
+{
+  template <typename Tuple>
+  _CCCL_HOST_DEVICE Tuple operator()(const Tuple& lhs, const Tuple& rhs)
+  {
+    using cuda::std::get;
+
+    return cuda::std::tuple(get<0>(lhs) + get<0>(rhs), get<1>(lhs) + get<1>(rhs));
+  }
+};
+
+struct MakeTupleFunctor
+{
+  template <typename T1, typename T2>
+  _CCCL_HOST_DEVICE cuda::std::tuple<T1, T2> operator()(T1& lhs, T2& rhs)
+  {
+    return cuda::std::tuple(lhs, rhs);
+  }
+};
+
+template <typename T>
+struct TestTupleReduce
+{
+  void operator()(const size_t n)
+  {
+    thrust::host_vector<T> h_t1 = random_integers<T>(n);
+    thrust::host_vector<T> h_t2 = random_integers<T>(n);
+
+    // zip up the data
+    thrust::host_vector<cuda::std::tuple<T, T>> h_tuples(n);
+    thrust::transform(h_t1.begin(), h_t1.end(), h_t2.begin(), h_tuples.begin(), MakeTupleFunctor());
+
+    // copy to device
+    thrust::device_vector<cuda::std::tuple<T, T>> d_tuples = h_tuples;
+
+    cuda::std::tuple<T, T> zero(0, 0);
+
+    // sum on host
+    cuda::std::tuple<T, T> h_result = thrust::reduce(h_tuples.begin(), h_tuples.end(), zero, SumTupleFunctor());
+
+    // sum on device
+    cuda::std::tuple<T, T> d_result = thrust::reduce(d_tuples.begin(), d_tuples.end(), zero, SumTupleFunctor());
+
+    ASSERT_EQUAL_QUIET(h_result, d_result);
+  }
+};
+VariableUnitTest<TestTupleReduce, IntegralTypes> TestTupleReduceInstance;