[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/cudax/test/stf/places/recursion.cu

@@ -0,0 +1,41 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of CUDASTF in CUDA C++ Core Libraries,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#include <cuda/experimental/stf.cuh>
+
+using namespace cuda::experimental::stf;
+
+void rec_func(exec_place places)
+{
+  if (places.size() == 1)
+  {
+    // places->print("SINGLE");
+  }
+  else
+  {
+    // places->print("REC");
+    for (int i = 0; i < 2; i++)
+    {
+      // Take every other places from the grid
+      auto half_places = partition_cyclic(places, dim4(2), pos4(i));
+      rec_func(half_places);
+    }
+  }
+}
+
+int main()
+{
+  auto places = exec_place::all_devices();
+  // places->print("ALL");
+
+  rec_func(places);
+
+  return 0;
+}