[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/examples/sum.cu

@@ -0,0 +1,39 @@
+#include <thrust/device_vector.h>
+#include <thrust/functional.h>
+#include <thrust/generate.h>
+#include <thrust/host_vector.h>
+#include <thrust/random.h>
+#include <thrust/reduce.h>
+
+#include <iostream>
+
+int my_rand()
+{
+  static thrust::default_random_engine rng;
+  static thrust::uniform_int_distribution<int> dist(0, 9999);
+  return dist(rng);
+}
+
+int main()
+{
+  // generate random data on the host
+  thrust::host_vector<int> h_vec(100);
+  thrust::generate(h_vec.begin(), h_vec.end(), my_rand);
+
+  // transfer to device and compute sum
+  thrust::device_vector<int> d_vec = h_vec;
+
+  // initial value of the reduction
+  int init = 0;
+
+  // binary operation used to reduce values
+  cuda::std::plus<int> binary_op;
+
+  // compute sum on the device
+  int sum = thrust::reduce(d_vec.begin(), d_vec.end(), init, binary_op);
+
+  // print the sum
+  std::cout << "sum is " << sum << '\n';
+
+  return 0;
+}