[BUG] Fix Segfault Caused by Incorrect Dereferencing of Renumber Map

[alexbarghi-nv]

Summary

• check the optional edge renumber map pointers before dereferencing them in
  the sampling-result C API accessors
• add regression coverage for accessing absent edge renumber maps on a
  renumbered homogeneous sampling result

Problem

cugraph_sample_result_get_edge_renumber_map() and
cugraph_sample_result_get_edge_renumber_map_offsets() currently check whether
the vertex renumber_map_ is null, then dereference the corresponding optional
edge renumber map pointer.

A sampling result can validly contain a vertex renumber map without an edge
renumber map. In particular, heterogeneous sampling with one edge type returns
that layout. Calling either accessor then dereferences a null pointer and
segfaults while pylibcugraph marshals an otherwise completed sampling result.

Fix

Each accessor now checks the pointer it actually dereferences:

• edge_renumber_map_
• edge_renumber_map_offsets_

The accessors return null for absent optional results, matching the behavior of
the other sampling-result accessors and the existing pylibcugraph null handling.

Validation

• repository pre-commit hooks pass for both changed files
• the cuGraph-GNN heterogeneous disjoint neighbor-loader reproducer passes
  end-to-end against an exact-baseline libcugraph build containing only this
  accessor fix
• the same reproducer still segfaults with cuGraph PR Fix bug found in how we handle empty frontiers in sampling #5629 alone, confirming
  that the empty-frontier fix is independent

[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.

[ChuckHastings]

/ok to test 3cae514

[ChuckHastings]

/merge