[codex] Add #1416 recursive ancestor row-join regression

[lmeyerov]

Summary

Adds focused regression coverage for #1416 / #880 around the GFQL recursive ancestor row-join lane.

The existing IS6-style test covered a Comment input where REPLY_OF*0.. reaches the Post ancestor through nonzero hops. This adds the matching Post input case, where the post ancestor must bind through the zero-hop arm before joining to forum/moderator rows.

Investigation

• The referenced rerun artifact results/runs/local-snb-interactive-closeout-2026-05-09 is not present in this checkout.
• In-repo residual triage only reads probe-results.json; the suite runner/adapter fallback producer is not present here.
• Current direct GFQL execution passes the nonzero-hop and zero-hop row-binding cases locally.

Validation

• python3 -m pytest -q graphistry/tests/test_compute_chain.py::TestChainBindingsTable::test_direct_rows_binding_ops_supports_open_range_multihop_continuation graphistry/tests/test_compute_chain.py::TestChainBindingsTable::test_direct_rows_binding_ops_supports_zero_hop_post_ancestor_join graphistry/tests/test_compute_chain.py::TestChainBindingsTable::test_direct_rows_binding_ops_supports_bounded_open_range_multihop_continuation
• ./bin/lint.sh graphistry/tests/test_compute_chain.py

Local blockers

• Full TestChainBindingsTable is not clean in this local environment: cuDF tests fail because no CUDA device is available, and row-expression tests fail because the parser backend dependency is missing.

Refs #1416. Fresh benchmark rerun confirmation is still required before closing the issue.