feat(vortex-bench): infra of SpatialBench on DuckDB

[HarukiMoriarty]

Summary

Adds a SpatialBench benchmark that runs on DuckDB across three storage lanes, Vortex native geometry, Vortex WKB, and GeoParquet. Spatial filters (ST_Distance, ST_Point, ST_X, ST_Y) lower into the Vortex scan and compute over columnar Point coordinates instead of materializing WKB each row.

Benchmark results

Local, warm, median of 5 (ms).

Q0 — radius filter (ST_Distance ≤ 50 km, ranked by distance)

Scale factor	vortex-native	vortex-WKB	GeoParquet	native vs WKB
SF1	6.4	15.5	41.4	2.4×
SF3	13.4	40.2	52.4	3.0×
SF10	32.8	129	189	3.9×

Q7 — building join (pickups within 500 m of each building)

Scale factor	vortex-native	vortex-WKB	GeoParquet
SF1	89	77	200
SF3	295	256	284
SF10	1053	928	1114

Takeaways

• Q0: native geometry pushdown wins decisively, the ST_Distance filter runs inside the scan over columnar f64 coordinates instead of materializing WKB per row.
• Q7: lanes converge — this is an engine-side spatial join where DuckDB builds a on-the-fly in-memory R-tree on the build side at query time, so it's format-blind. The native lane even pays a small f64→WKB serialization cost to feed the join.

[codspeed-hq]

Merging this PR will not alter performance

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 2 improved benchmarks
❌ 3 regressed benchmarks
✅ 1578 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	chunked_varbinview_canonical_into[(1000, 10)]	154.6 µs	190.9 µs	-19.02%
❌	Simulation	bitwise_not_vortex_buffer_mut[128]	215.3 ns	244.4 ns	-11.93%
❌	Simulation	rebuild_naive	99.8 µs	111 µs	-10.09%
⚡	Simulation	chunked_bool_canonical_into[(1000, 10)]	26.4 µs	16.1 µs	+63.33%
⚡	Simulation	chunked_varbinview_canonical_into[(100, 100)]	260.4 µs	223.7 µs	+16.42%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing nemo/spatialbench-duckdb (4216cd0) with develop (e6d0e08)}