[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/buffer/gdb.expected

@@ -0,0 +1,63 @@
+=============== buffer.normal begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=10, align=4, data=<address> (device) = {
+  [0] = -56,
+  [1] = 22,
+  [2] = 94,
+  [3] = -13,
+  [4] = 7,
+  [5] = 41,
+  [6] = -82,
+  [7] = 0,
+  [8] = 63,
+  [9] = -5
+}
+=============== buffer.normal end ===============
+=============== buffer.alias begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=4, align=4, data=<address> (device) = {
+  [0] = 17,
+  [1] = -31,
+  [2] = 8,
+  [3] = 55
+}
+=============== buffer.alias end ===============
+=============== buffer.vector.0 begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=3, align=4, data=<address> (device) = {
+  [0] = -2,
+  [1] = 4,
+  [2] = 6
+}
+=============== buffer.vector.0 end ===============
+=============== buffer.vector.1 begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=3, align=4, data=<address> (device) = {
+  [0] = 11,
+  [1] = -9,
+  [2] = 27
+}
+=============== buffer.vector.1 end ===============
+=============== buffer.host_device begin ===============
+cuda::buffer<int, cuda::mr::device_accessible, cuda::mr::host_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible, cuda::mr::host_accessible> @ <address>, stream=<address>, size=4, align=4, data=<address> (host/device) = {
+  [0] = 3,
+  [1] = 14,
+  [2] = -15,
+  [3] = 92
+}
+=============== buffer.host_device end ===============
+=============== buffer.empty begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=0, align=4, data=0x0 (device)
+=============== buffer.empty end ===============
+=============== buffer.update.before begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=4, align=4, data=<address> (device) = {
+  [0] = 1,
+  [1] = 2,
+  [2] = 3,
+  [3] = 4
+}
+=============== buffer.update.before end ===============
+=============== buffer.update.after begin ===============
+cuda::buffer<int, cuda::mr::device_accessible> mr=cuda::mr::any_resource<cuda::mr::device_accessible> @ <address>, stream=<address>, size=4, align=4, data=<address> (device) = {
+  [0] = -8,
+  [1] = 13,
+  [2] = 21,
+  [3] = -34
+}
+=============== buffer.update.after end ===============