test(gfql): shape registry fed by the specialization tables + route harness

[lmeyerov]

Stacked on #2054. The route harness the review of #2054 asked for: one registered input is tried against every chain route whose admission predicate admits it, so the specializations' own test tables amplify each other.

Registry (graphistry/tests/compute/gfql/routes/registry.py): each specialization's test module registers the shape table it already owns with its frames and defect-class tags; register(...) returns the table unchanged. Registered today: the routes corpus (22), the six test_chain.py tables (44), the alias-collision matrix (14): 80 shapes.

Harness (tests/compute/gfql/routes/test_route_harness.py), per admitted cell (native-fast on pandas/cuDF, polars-plain, polars-seeded): the lane serves; the answer matches the same engine's general path on node/edge values (every route declined through the shared routes_off switch); node/edge sets match the pandas general path. A lane that declines an admitted shape is recorded as an expected failure (the attenuation ledger: the polars seeded lane declines the 10 alias-collision shapes its predicate admits → #2063, measured at 29 ms vs 1.9 ms pandas on the SNB sentinel); a serve-ratio pin fails when a lane declines most of what it admits. Filed divergences are strict expected failures keyed by tag; the pandas _BYPASS_SHAPES prune rows surfaced #2053 on the polars route this way (fixed in #2065, stacked above).

Engagement marks. The 7-mode routes-off replay (ledger table on #2054, files under reviews/2054/) classified every remaining failing id; the tests that assert a route serves now carry @pytest.mark.route_engaged(...), so the replay reports result divergences only. What remained were #2058 (fixed in #2062), #2059 (typed decline) and #2034 (contract question posted).

Local: 382 pass / 13 xfail on routes + test_chain + collision (pandas, cuDF, polars); broad CPU 12592 pass; cuDF 1390 pass. GPU receipt on the stack head below. tck-gfql per-route ledger: graphistry/tck-gfql#196.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QztW7jYsDd66e8rb8pJNQA

[lmeyerov]

GPU receipt (dgx-spark, graphistry/test-rapids-official:26.02-gfql-polars, cudf 26.02.01 / cupy 13.6.0 / 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 stack head 676232a (#2062, which contains #2055, #2056, #2054 and #2061): 13383 passed / 93 failed / 55 skipped / 44 xfailed / 4 xpassed. The 93 failures are exactly master 86de0f5's set in the same image (13227 passed / 93 failed: the polars-gpu fused grouped-aggregate lane and strictness tests, #2064): set difference empty, zero branch-specific failures.

[lmeyerov]

CI receipt at 21f662c: 82 check-runs, all success (incl. tck-gfql, test-polars ×6 with the harness in the lane and the coverage audit, gfql-benchmarks, the non-blocking gfql-routes-off ledger matrix). GPU receipt on the stack head above. Marking ready.

[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 00002b8 (restacked on the #2055 rework): 82 check-runs, all success. GPU receipt on the stack head above.