Migrate RMM usage to CCCL MR design

[bdice]

Summary

• Replace removed rmm::mr::device_memory_resource base class, owning_wrapper, shared_ptr-based resource management, and deprecated per-device resource APIs with CCCL-native memory resource types
• Use cuda::mr::any_resource<cuda::mr::device_accessible> for owning type-erased storage, rmm::device_async_resource_ref for non-owning references, and value-typed resources (cuda_memory_resource, pinned_host_memory_resource)
• Remove invalid 3rd MR argument from raft::handle_t constructor calls (handle only accepts stream + stream_pool)

Files changed (15)

Headers:

• algorithms.hpp, dendrogram.hpp, legacy/graph.hpp, legacy/functions.hpp: get_current_device_resource() → get_current_device_resource_ref()
• host_staging_buffer_manager.hpp: Remove owning_wrapper, use pool_memory_resource directly via std::optional, accept pinned_host_memory_resource by value
• large_buffer_manager.hpp: Store pinned_host_memory_resource by value (not shared_ptr), return device_async_resource_ref from get()
• mtmg/resource_manager.hpp: Use cuda::mr::any_resource<device_accessible> instead of shared_ptr<device_memory_resource>, use set_per_device_resource_ref, remove MR arg from raft::handle_t

Tests:

• base_fixture.hpp: Return any_resource from create_memory_resource(), use value-typed resources, set_current_device_resource_ref
• multi_node_threaded_test.cpp: set_current_device_resource_ref(resource)
• mg_graph500_bfs_test.cu: pinned_mr_ as optional<pinned_host_memory_resource> value type
• mg_graph500_sssp_test.cu: pinned_memory_resource → pinned_host_memory_resource value type

Examples:

• All 4 example files: value-typed cuda_memory_resource, set_current_device_resource_ref, remove MR from handle constructor

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.