fix(gfql): one op-list alias-shadowing contract on every engine (#2049, #2039 polars edge side)

[lmeyerov]

Closes #2049. Two commits.

Contract (the surface decides, not the engine). Op-list chains: an alias named like a column of the frame it marks becomes the boolean marker under that name, on pandas, cuDF and polars alike (shadowing). Cypher: the variable keeps reading the user value through its properties on every engine (MATCH (a)-[type:K]->(b) RETURN type.type answers the user values), because the shadowed values ride under the internal restore column that the row pipeline already resolves.

Commit 1 (eager multi-hop, #2049). A non-simple step (hops=2, to_fixed_point) whose edge alias names the column its own edge_match filters on had the chain's backward pass re-read the marker instead of the user values; the backward re-execution now runs on the graph's own edge rows for the step's ids (_step_with_source_edge_columns).

Commit 2 (polars edge side). The polars chain kept the user values under the alias and leaked the marker as a type_right join suffix (_combine_edges joined the marker frame onto the original edges). Now the marker takes the name and the user column moves to shadow_restore_column(alias); the polars rows-route resolver and the binding-rows payload prefer that column for alias.alias, and the rows-route restore short-circuits when it is present. No _right columns.

Pins. test_chain_alias_column_collision.py: every served single-hop shape and both multi-hop shapes compare the marker values across pandas/cuDF/polars and reject join suffixes. .../polars/test_chain_alias_column_collision_2039.py: single-hop and two-hop edge alias marker equals pandas'. test_alias_scoping_semantics.py: the polars residual (type.type raised a typed decline) flips to parity with pandas; the other 14 scoping pins unchanged (Cypher reads user values on both engines).

Visible change. On polars, an op-list result whose alias shadowed an edge column now carries the marker under the alias and the user values under __gfql_shadow_restore__<alias>__ instead of <alias> + <alias>_right. Stated in CHANGELOG.

Receipts: collision + 2039 + scoping + polars seed-resolution suites 222 pass / 9 xfail (cuDF on); guards + mypy green; broad suites running, dgx lane on the stack head to follow.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QztW7jYsDd66e8rb8pJNQA

[lmeyerov]

READY (3312f00, on master b8f3e99) — CI 69/69 non-skipped check-runs success; the 8 skipped lanes are the gfql path filter (chain.py is outside it; #2057 proposes the fix). tck-gfql run locally from its current main (66fee10, the ref CI would use) against this head: exit 0 — 3658 scenarios, 2964 supported, 1977 translated-supported, 689 xfail (baseline), direct-Cypher non-validation debt 0, no outcome drift. Local suites: 606 passed / 9 xfailed across collision matrix, chain, hop, alias-scoping, residual-polars, hop-scaling and native-seed with the cuDF lane; guards OK; mypy = master; ruff clean.

[lmeyerov]

GPU receipt (dgx-spark, graphistry/test-rapids-official:26.02-gfql-polars, cudf 26.02.01 / polars 1.35.2, safe_run, TEST_CUDF=1; suites test_chain.py, test_hop.py, test_chain_alias_column_collision.py, tests/compute/gfql/**) on the restacked stack head a54ac4b (#2065, containing #2055 rework, #2056, #2054, #2061, #2062): 13386 passed / 95 failed / 38 xfailed / 4 xpassed. 93 of the failures are master 86de0f5's polars-gpu fused-lane set (#2064, fixed separately in #2066). The other 2 are strict XPASS of the #2043 prune_to_endpoints markers on cuDF 26.02: the prune shapes now agree fast-vs-full on this stack (the #2062 closure fix removed the divergence), so #2062 drops that marker in a follow-up commit (tree otherwise identical; #2065 restacked).

[lmeyerov]

CI receipt at 510dcc6 (rework with the unified alias-shadowing contract): 72 check-runs success, 5 path-skipped as whole jobs (tck-gfql, test-polars ×6, test-gfql-core, gfql-benchmarks and the guards green). Broad local suites 12557 pass; GPU receipt on the stack head above.