[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/libcudacxx/test/debugging/array/CMakeLists.txt

@@ -0,0 +1,13 @@
+libcudacxx_add_pretty_printer_test(
+  NAME array
+  SOURCES source.cu
+  CASES
+    # gersemi: off
+    --case inspect_normal 1 array.normal normal
+    --case inspect_empty 1 array.empty empty
+    --case inspect_nested 1 array.nested nested
+    --case inspect_alias 1 array.alias alias
+    --case inspect_before_update 1 array.update.before updated_values
+    --case inspect_after_update 1 array.update.after updated_values
+    # gersemi: on
+)

cccl_upstream/libcudacxx/test/debugging/array/gdb.expected

@@ -0,0 +1,44 @@
+=============== array.normal begin ===============
+cuda::std::array<int, 3> = {
+  [0] = -7,
+  [1] = 0,
+  [2] = 42
+}
+=============== array.normal end ===============
+=============== array.empty begin ===============
+cuda::std::array<int, 0>
+=============== array.empty end ===============
+=============== array.nested begin ===============
+cuda::std::array<cuda::std::array<int, 2>, 2> = {
+  [0] = cuda::std::array<int, 2> = {
+    [0] = 13,
+    [1] = -5
+  },
+  [1] = cuda::std::array<int, 2> = {
+    [0] = 0,
+    [1] = 88
+  }
+}
+=============== array.nested end ===============
+=============== array.alias begin ===============
+cuda::std::array<int, 4> = {
+  [0] = -31,
+  [1] = 17,
+  [2] = 8,
+  [3] = -64
+}
+=============== array.alias end ===============
+=============== array.update.before begin ===============
+cuda::std::array<int, 3> = {
+  [0] = 6,
+  [1] = -91,
+  [2] = 52
+}
+=============== array.update.before end ===============
+=============== array.update.after begin ===============
+cuda::std::array<int, 3> = {
+  [0] = 3,
+  [1] = 85,
+  [2] = -12
+}
+=============== array.update.after end ===============

cccl_upstream/libcudacxx/test/debugging/array/lldb.expected

@@ -0,0 +1,21 @@
+=============== array.normal begin ===============
+(cuda::std::array<int, 3>)  ([0] = -7, [1] = 0, [2] = 42)
+=============== array.normal end ===============
+=============== array.empty begin ===============
+(cuda::std::array<int, 0>)
+=============== array.empty end ===============
+=============== array.nested begin ===============
+(cuda::std::array<cuda::std::array<int, 2>, 2>) {
+  [0] = ([0] = 13, [1] = -5)
+  [1] = ([0] = 0, [1] = 88)
+}
+=============== array.nested end ===============
+=============== array.alias begin ===============
+(cuda::std::array<int, 4>)  ([0] = -31, [1] = 17, [2] = 8, [3] = -64)
+=============== array.alias end ===============
+=============== array.update.before begin ===============
+(cuda::std::array<int, 3>)  ([0] = 6, [1] = -91, [2] = 52)
+=============== array.update.before end ===============
+=============== array.update.after begin ===============
+(cuda::std::array<int, 3>)  ([0] = 3, [1] = 85, [2] = -12)
+=============== array.update.after end ===============

cccl_upstream/libcudacxx/test/debugging/array/source.cu

@@ -0,0 +1,60 @@
+// Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+//
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <cuda/std/array>
+
+template <class T>
+[[gnu::noinline]] void keep_for_debugger(const T& value)
+{
+  asm volatile("" : : "g"(&value) : "memory");
+}
+
+using array_alias = cuda::std::array<int, 4>;
+
+[[gnu::noinline]] void inspect_normal(const cuda::std::array<int, 3>& values)
+{
+  keep_for_debugger(values);
+}
+
+[[gnu::noinline]] void inspect_empty(const cuda::std::array<int, 0>& values)
+{
+  keep_for_debugger(values);
+}
+
+[[gnu::noinline]] void inspect_nested(const cuda::std::array<cuda::std::array<int, 2>, 2>& values)
+{
+  keep_for_debugger(values);
+}
+
+[[gnu::noinline]] void inspect_alias(const array_alias& values)
+{
+  keep_for_debugger(values);
+}
+
+[[gnu::noinline]] void inspect_before_update(const cuda::std::array<int, 3>& values)
+{
+  keep_for_debugger(values);
+}
+
+[[gnu::noinline]] void inspect_after_update(const cuda::std::array<int, 3>& values)
+{
+  keep_for_debugger(values);
+}
+
+int main()
+{
+  const cuda::std::array<int, 3> normal                      = {-7, 0, 42};
+  const cuda::std::array<int, 0> empty                       = {};
+  const cuda::std::array<cuda::std::array<int, 2>, 2> nested = {{{13, -5}, {0, 88}}};
+  const array_alias alias                                    = {-31, 17, 8, -64};
+  cuda::std::array<int, 3> updated_values                    = {6, -91, 52};
+
+  inspect_normal(normal);
+  inspect_empty(empty);
+  inspect_nested(nested);
+  inspect_alias(alias);
+  inspect_before_update(updated_values);
+  updated_values = {3, 85, -12};
+  inspect_after_update(updated_values);
+}