[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/counting_iterator.cu

@@ -0,0 +1,34 @@
+#include <thrust/copy.h>
+#include <thrust/device_vector.h>
+#include <thrust/functional.h>
+#include <thrust/iterator/counting_iterator.h>
+
+#include <iostream>
+#include <iterator>
+
+int main()
+{
+  // this example computes indices for all the nonzero values in a sequence
+
+  // sequence of zero and nonzero values
+  thrust::device_vector<int> stencil{0, 1, 1, 0, 0, 1, 0, 1};
+
+  // storage for the nonzero indices
+  thrust::device_vector<int> indices(8);
+
+  // counting iterators define a sequence [0, 8)
+  thrust::counting_iterator<int> first(0);
+  thrust::counting_iterator<int> last = first + 8;
+
+  // compute indices of nonzero elements
+  using IndexIterator = thrust::device_vector<int>::iterator;
+
+  IndexIterator indices_end = thrust::copy_if(first, last, stencil.begin(), indices.begin(), cuda::std::identity{});
+  // indices now contains [1,2,5,7]
+
+  // print result
+  std::cout << "found " << cuda::std::distance(indices.begin(), indices_end) << " nonzero values at indices:\n";
+  thrust::copy(indices.begin(), indices_end, std::ostream_iterator<int>(std::cout, "\n"));
+
+  return 0;
+}