[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/memory_resource/CMakeLists.txt

@@ -0,0 +1,10 @@
+libcudacxx_add_pretty_printer_test(
+  NAME memory_resource
+  SOURCES source.cu
+  CASES
+    # gersemi: off
+    --case inspect_device 1 memory_resource.device device_resource
+    --case inspect_host_device 1 memory_resource.host_device host_device_resource
+    --case inspect_alias 1 memory_resource.alias aliased_resource
+    # gersemi: on
+)

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

@@ -0,0 +1,9 @@
+=============== memory_resource.device begin ===============
+cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>
+=============== memory_resource.device end ===============
+=============== memory_resource.host_device begin ===============
+cuda::mr::any_resource<cuda::mr::device_accessible, cuda::mr::host_accessible> @ <address>
+=============== memory_resource.host_device end ===============
+=============== memory_resource.alias begin ===============
+cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>
+=============== memory_resource.alias end ===============

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

@@ -0,0 +1,9 @@
+=============== memory_resource.device begin ===============
+(const device_resource_type) cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>
+=============== memory_resource.device end ===============
+=============== memory_resource.host_device begin ===============
+(const host_device_resource_type) cuda::mr::any_resource<cuda::mr::device_accessible, cuda::mr::host_accessible> @ <address>
+=============== memory_resource.host_device end ===============
+=============== memory_resource.alias begin ===============
+(const resource_alias) cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>
+=============== memory_resource.alias end ===============

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

@@ -0,0 +1,43 @@
+// Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+//
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <cuda/memory_resource>
+
+template <class T>
+[[gnu::noinline]] void keep_for_debugger(const T& value)
+{
+  asm volatile("" : : "g"(&value) : "memory");
+}
+
+using device_resource_type      = cuda::mr::any_resource<cuda::mr::device_accessible>;
+using host_device_resource_type = cuda::mr::any_resource<cuda::mr::device_accessible, cuda::mr::host_accessible>;
+using resource_alias            = device_resource_type;
+
+[[gnu::noinline]] void inspect_device(const device_resource_type& resource)
+{
+  keep_for_debugger(resource);
+}
+
+[[gnu::noinline]] void inspect_host_device(const host_device_resource_type& resource)
+{
+  keep_for_debugger(resource);
+}
+
+[[gnu::noinline]] void inspect_alias(const resource_alias& resource)
+{
+  keep_for_debugger(resource);
+}
+
+int main()
+{
+  using adapted_resource = cuda::mr::synchronous_resource_adapter<cuda::mr::legacy_managed_memory_resource>;
+  const adapted_resource managed_resource{cuda::mr::legacy_managed_memory_resource{}};
+  const device_resource_type device_resource{managed_resource};
+  const host_device_resource_type host_device_resource{managed_resource};
+  const resource_alias aliased_resource{managed_resource};
+
+  inspect_device(device_resource);
+  inspect_host_device(host_device_resource);
+  inspect_alias(aliased_resource);
+}