GFQL/Cypher: drive down expr-temporal wrong-row tranche (14 cases, #1353 item #2) #1361

Summary

Concrete tranche from meta-issue #1353 item #2 — Temporal comparison / canonicalization. Drives down 14 `success_wrong_rows` cases under `expressions/temporal` to `success_matches_expected` without growing `unexpected_success_expected_error`.

Targets:

• `expr-temporal2-6-5`
• `expr-temporal6-6-2`, `expr-temporal6-6-8`
• `expr-temporal7-1-1`, `expr-temporal7-1-2`
• `expr-temporal7-2-1`, `expr-temporal7-2-2`
• `expr-temporal7-3-1`, `expr-temporal7-3-2`
• `expr-temporal7-4-1`, `expr-temporal7-4-2`
• `expr-temporal7-5-1`, `expr-temporal7-5-2`
• `expr-temporal7-6-8`

Approach

1. Triage first — for each scenario, capture: actual rows, expected rows, root-cause hypothesis (compile vs. row-pipeline temporal comparator vs. fixture-parser limitation), candidate fix locus. Findings live under `plans/1361-temporal-wrong-rows/findings/` (gitignored).
2. Decide ordering after triage. Prefer one issue / one branch / one or two clean commits over a PR stack to limit management overhead.
3. Implement fixes with behavioral tests under `graphistry/tests/compute/gfql/...` and contract update in tck-gfql `tests/cypher_tck/direct_cypher_xfail_contract.py`.
4. Use repo-local review skill (`agents/skills/review/SKILL.md`); iterate until convergence.
5. cuDF dual-version validation on `dgx-spark` (RAPIDS 25.02 + 26.02) only if changes touch row-pipeline runtime path; skip for compile-time-only.

Lesson from #1359

Item #1's 14 wrong-row cases all bottomed out as tck-gfql repo bugs (parser stored list literals as strings; `_rows_ordered` ignored "in any order"). Tracked in tck-gfql#36. Temporal values survive `_parse_properties` cleanly (string literal in CREATE → string in df, which GFQL handles natively), so this tranche is more likely to contain real pygraphistry semantic bugs.

The `_rows_ordered` fix in tck-gfql#36 will likely incidentally clear some of these scenarios too — re-snapshot after that lands.

Acceptance criteria

• 14 listed scenarios move from `success_wrong_rows` → `success_matches_expected`.
• `unexpected_success_expected_error` does not increase.
• `tck-gfql` focused contract checks remain green.
• Existing `graphistry/tests/compute/gfql/...` suite remains green.

Repro

```bash
PYTHONPATH=/home/lmeyerov/Work/pygraphistry:/home/lmeyerov/Work/tck-gfql UV_EXCLUDE_NEWER="6 days" \
uv run --python 3.13 --no-project --with pytest --with pandas \
python -m pytest -q tests/cypher_tck/test_tck_runner.py -k "test_direct_cypher_xfail_contract"
```

Cross-reference

Tracks meta-issue #1353 item #2. Sibling: #1359 (item #1, gated on tck-gfql#36).