[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/benchmarks/bench/extrema/basic.cu

@@ -0,0 +1,54 @@
+// SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <thrust/device_vector.h>
+#include <thrust/execution_policy.h>
+#include <thrust/extrema.h>
+
+#include "nvbench_helper.cuh"
+
+template <typename T, typename Func>
+static void bench_extremum(nvbench::state& state, nvbench::type_list<T>, Func func)
+{
+  const auto elements = static_cast<std::size_t>(state.get_int64("Elements"));
+
+  thrust::device_vector<T> in = generate(elements);
+
+  using offset_t = typename decltype(in.cbegin())::difference_type;
+
+  state.add_element_count(elements);
+  state.add_global_memory_reads<T>(elements);
+  state.add_global_memory_writes<offset_t>(1);
+
+  caching_allocator_t alloc;
+  state.exec(nvbench::exec_tag::gpu | nvbench::exec_tag::no_batch | nvbench::exec_tag::sync,
+             [&](nvbench::launch& launch) {
+               do_not_optimize(func(policy(alloc, launch), in.cbegin(), in.cend()));
+             });
+}
+
+template <typename T>
+static void min_element(nvbench::state& state, nvbench::type_list<T> list)
+{
+  bench_extremum(state, list, [](auto&&... args) {
+    return thrust::min_element(args...);
+  });
+}
+
+NVBENCH_BENCH_TYPES(min_element, NVBENCH_TYPE_AXES(fundamental_types))
+  .set_name("min_element")
+  .set_type_axes_names({"T{ct}"})
+  .add_int64_power_of_two_axis("Elements", nvbench::range(16, 28, 4));
+
+template <typename T>
+static void max_element(nvbench::state& state, nvbench::type_list<T> list)
+{
+  bench_extremum(state, list, [](auto&&... args) {
+    return thrust::max_element(args...);
+  });
+}
+
+NVBENCH_BENCH_TYPES(max_element, NVBENCH_TYPE_AXES(fundamental_types))
+  .set_name("max_element")
+  .set_type_axes_names({"T{ct}"})
+  .add_int64_power_of_two_axis("Elements", nvbench::range(16, 28, 4));