[codex] Fix Cypher zero-hop and empty post-aggregate rows

[lmeyerov]

Summary

• Fix exact zero-hop Cypher variable-length relationships so *0..0 preserves the seed node as the zero-length path while keeping the public hop(..., 0) API behavior unchanged.
• Fix empty-graph post-aggregate expression projection so MATCH (a) RETURN count(a) > 0 synthesizes the final output column/value.
• Fix amplification-discovered sibling post-aggregate expression temp-name collisions such as RETURN count(a) = 0, count(a) + 1.
• Add Tranche: mixed wrong-row tail cleanup under #1353 #1369 changelog, pandas/cuDF-focused regressions, intentional Cypher surface-guard baseline update, and same-named tck-gfql companion contract branch.

Refs #1369. Refs #1353.

Validation

• /tmp/tck-gfql-ci-venv312/bin/python -m pytest -q graphistry/tests/test_compute_hops.py::TestComputeHopMixin::test_hop_0b -> 1 passed.
• /tmp/tck-gfql-ci-venv312/bin/python -m pytest -q graphistry/tests/compute/gfql/cypher/test_lowering.py -k "issue_1369" -> 4 passed, 2 local cuDF skips, 1019 deselected, 4 warnings.
• /tmp/tck-gfql-ci-venv312/bin/python -m pytest -q graphistry/tests/test_compute_hops.py::TestComputeHopMixin::test_hop_0 graphistry/tests/test_compute_hops.py::TestComputeHopMixin::test_hop_0b graphistry/tests/compute/gfql/cypher/test_lowering.py -k "issue_1369 or issue_983 or two_match_reentry_varlen" -> 12 passed, 2 local cuDF skips, 1013 deselected, 4 warnings.
• ./bin/ruff.sh graphistry/Plottable.py graphistry/compute/ast.py graphistry/compute/hop.py graphistry/compute/gfql/cypher/lowering.py graphistry/tests/compute/gfql/cypher/test_lowering.py -> passed.
• /bin/bash -lc "source /tmp/tck-gfql-ci-venv312/bin/activate && ./bin/typecheck.sh" -> passed.
• python3 bin/ci_cypher_surface_guard.py -> passed after intentional lowering.py baseline bump to 9022 lines for the empty post-aggregate projection + amplification fix.
• DGX RAPIDS 25.02: RAPIDS_VERSION=25.02 PROFILE=gfql WITH_GPU=1 ... TEST_FILES="<two #1369 cuDF tests>" docker/test-rapids-official-local.sh -> cudf 25.02.02 import smoke; 2 passed.
• DGX RAPIDS 26.02: RAPIDS_VERSION=26.02 PROFILE=gfql WITH_GPU=1 ... TEST_FILES="<two #1369 cuDF tests>" docker/test-rapids-official-local.sh -> cudf 26.02.01 import smoke; 2 passed.
• PYTHONPATH=/home/lmeyerov/Work/pygraphistry4:/tmp/tck-gfql-issue-1369 /tmp/tck-gfql-ci-venv312/bin/python -m pytest -q tests/cypher_tck/test_direct_cypher_contract_fastfail.py tests/cypher_tck/test_report.py -> 8 passed.
• PYTHONPATH=/home/lmeyerov/Work/pygraphistry4:/tmp/tck-gfql-issue-1369 PYGRAPHISTRY_PATH=/home/lmeyerov/Work/pygraphistry4 PYGRAPHISTRY_INSTALL=0 ./bin/ci.sh in sibling tck-gfql worktree -> 3935 passed, 766 xfailed.
• Companion graphistry/tck-gfql@issue-1369-mixed-wrong-row-tail is rebased/pushed at bd9107f.
• Review skill reconverged after final amplification delta: 6 waves total, 0 blocker / 0 important / 0 suggestions.
• Rebased onto origin/master @ 79d62aa5 after conflict resolution; pushed head is 90aebc74.
• Post-rebase local checks: git diff --check, focused Tranche: mixed wrong-row tail cleanup under #1353 #1369 pytest, ruff touched files, Cypher surface guard, and broader hop/Cypher cluster all passed.
• Post-rebase DGX RAPIDS 25.02/26.02 cuDF reruns passed (2 passed on each).
• Current PR CI for pushed 90aebc74 is green; merge remains blocked only on required GitHub review.

Residual #1369 notes

• Fixed/promoted by this lane: match5-8 and return2-10 now return success_matches_expected; expr-aggregation3-1 and expr-mathematical8-1/2 were already supported/matches-expected on the current sibling snapshot.
• Residual wrong-row keys remain: expr-pattern1-10, match5-25, match5-26, return2-9, and with2-1.
• with2-1 appears blocked by sibling fixture/port behavior: the graph fixture materializes (:Begin {num: a.id}) as string "a.id", so the expected join value is not available to pygraphistry runtime.
• return2-9 expects duplicate standalone map rows despite current one-row RETURN semantics; likely needs TCK contract/fixture follow-up.