feat(cccl): integrate missing CCCL directories — python/, ci/, .agent/, docs/, test/

Sparse-checkout from NVIDIA/cccl main branch to complete cccl_upstream:

Added:
- python/cuda_cccl/ (226 files) — Python bindings for device-level algorithms
  Critical for muh toolchain: cuda.compute.reduce_into, scan, radix_sort, etc.
  Includes 204 .py files with full test coverage for all 27 algorithms
- ci/ (163 files) — Build/test infrastructure
  build_cub.sh, test_cub.sh, build_and_test_targets.sh, matrix.yaml
  Directly maps to our [INFRA-CI] and [INFRA-BUILD] items
- .agent/skills/ (7 files) — NVIDIA's own agent skills for CCCL
  cccl-style/SKILL.md, cccl-test/SKILL.md, sass-diff/SKILL.md
- docs/ (491 files) — Official CCCL documentation
  CI references, CMake guides, Python compute docs, libcudacxx PTX docs
- test/ (12 files) — Top-level integration tests (cuda_smoke, stdpar)
- Root configs: .clang-format, .clang-tidy, CONTRIBUTING.md, pyproject.toml
- CLAUDE.md symlink → AGENTS.md (NVIDIA's standard)

cccl_upstream now mirrors full NVIDIA/cccl structure:
  Before: 42M (cub + thrust + libcudacxx + cudax + c + examples + benchmarks)
  After:  53M (+python +ci +docs +.agent +test +configs)

This completes the CCCL base needed for:
- [muh-bench] items: ci/util/build_and_test_targets.sh for targeted builds
- [CCCL-verify] items: python/cuda_cccl/tests/ as reference implementations
- [CCCL-test] items: ci/test_cub.sh, ci/test_thrust.sh
- Agent workflow: .agent/skills/ for consistent style and test patterns

cccl_upstream/ci/build_cub.sh

@@ -0,0 +1,111 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+NO_LID=false
+LID0=false
+LID1=false
+LID2=false
+ARTIFACT_TAGS=()
+
+ci_dir=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+
+new_args="$("${ci_dir}/util/extract_switches.sh" \
+  -no-lid \
+  -lid0 \
+  -lid1 \
+  -lid2 \
+  -- "$@")"
+
+declare -a new_args="(${new_args})"
+set -- "${new_args[@]}"
+while true; do
+  case "$1" in
+  -no-lid)
+    ARTIFACT_TAGS+=("no_lid")
+    NO_LID=true
+    shift
+    ;;
+  -lid0)
+    ARTIFACT_TAGS+=("lid_0")
+    LID0=true
+    shift
+    ;;
+  -lid1)
+    ARTIFACT_TAGS+=("lid_1")
+    LID1=true
+    shift
+    ;;
+  -lid2)
+    ARTIFACT_TAGS+=("lid_2")
+    LID2=true
+    shift
+    ;;
+  --)
+    shift
+    break
+    ;;
+  *)
+    echo "Unknown argument: $1"
+    exit 1
+    ;;
+  esac
+done
+
+# shellcheck source=ci/build_common.sh
+source "${ci_dir}/build_common.sh"
+
+print_environment_details
+
+ENABLE_CCCL_BENCHMARKS="false"
+ENABLE_CUB_RDC="false"
+
+if [[ "$CUDA_COMPILER" == *nvcc* ]]; then
+    ENABLE_CUB_RDC="true"
+    NVCC_VERSION=$($CUDA_COMPILER --version | grep release | awk '{print $6}' | cut -c2-)
+    if [[ -n "${DISABLE_CUB_BENCHMARKS}" ]]; then
+        echo "Benchmarks have been forcefully disabled."
+    else
+        ENABLE_CCCL_BENCHMARKS="true"
+        echo "nvcc version is $NVCC_VERSION. Building CUB benchmarks."
+    fi
+else
+    echo "Not building with NVCC, disabling RDC and benchmarks."
+fi
+
+if [[ "$HOST_COMPILER" == *icpc* || "$HOST_COMPILER" == *nvhpc* ]]; then
+    ENABLE_CCCL_BENCHMARKS="false"
+fi
+
+PRESET="cub"
+if $NO_LID; then
+    PRESET="cub-nolid"
+elif $LID0; then
+    PRESET="cub-lid0"
+elif $LID1; then
+    PRESET="cub-lid1"
+elif $LID2; then
+    PRESET="cub-lid2"
+fi
+
+CMAKE_OPTIONS=(
+    "-DCMAKE_CXX_STANDARD=$CXX_STANDARD"
+    "-DCMAKE_CUDA_STANDARD=$CXX_STANDARD"
+    "-DCCCL_ENABLE_BENCHMARKS=$ENABLE_CCCL_BENCHMARKS"
+    "-DCUB_ENABLE_RDC_TESTS=$ENABLE_CUB_RDC"
+)
+
+configure_and_build_preset "CUB" "$PRESET" "${CMAKE_OPTIONS[@]}"
+
+# Create test artifacts:
+if [[ -n "${GITHUB_ACTIONS:-}" ]]; then
+    if [[ ${#ARTIFACT_TAGS[@]} -gt 0 ]]; then
+        run_command "📦  Packaging test artifacts" \
+            "${ci_dir}/upload_cub_test_artifacts.sh" \
+            "${ARTIFACT_TAGS[@]}"
+    else
+        run_command "📦  Packaging test artifacts" "${ci_dir}/upload_cub_test_artifacts.sh"
+    fi
+fi
+
+print_time_summary