[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/max_abs_diff.cu

@@ -0,0 +1,37 @@
+#include <thrust/device_vector.h>
+#include <thrust/functional.h>
+#include <thrust/inner_product.h>
+
+#include <cuda/functional>
+
+#include <iostream>
+
+// this example computes the maximum absolute difference
+// between the elements of two vectors
+
+template <typename T>
+struct abs_diff
+{
+  __host__ __device__ T operator()(const T& a, const T& b)
+  {
+    return fabsf(b - a);
+  }
+};
+
+int main()
+{
+  thrust::device_vector<float> d_a = {1.0, 2.0, 3.0, 4.0};
+  thrust::device_vector<float> d_b = {2.0, 4.0, 3.0, 0.0};
+
+  // initial value of the reduction
+  float init = 0;
+
+  // binary operations
+  cuda::maximum<float> binary_op1{};
+  abs_diff<float> binary_op2;
+
+  float max_abs_diff = thrust::inner_product(d_a.begin(), d_a.end(), d_b.begin(), init, binary_op1, binary_op2);
+
+  std::cout << "maximum absolute difference: " << max_abs_diff << '\n';
+  return 0;
+}