feat(vortex-spatial): add convex hull scalar function

[HarukiMoriarty]

Summary

• add a strict native MultiPoint to Polygon convex-hull scalar function
• delegate planar hull computation directly to geo::ConvexHull
• preserve CRS and outer nullability while materializing native 2-D Polygon storage
• use shared unary dispatch for constants and strict null propagation
• cover empty, one-point, collinear, nullable, constant, and Q5-style collect-to-hull-to-area execution
• add a dedicated convex-hull microbenchmark

Semantics

The return dtype is always native Polygon, including empty and degenerate hulls. Consistent with the current geometry-kernel policy, coordinate finiteness is unchecked unless validation is explicitly enabled; this function does not add per-call NaN validation.

[codspeed-hq]

Merging this PR will regress 1 benchmark

⚡ 23 improved benchmarks
❌ 1 regressed benchmark
✅ 1925 untouched benchmarks
🆕 18 new benchmarks
⏩ 85 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	cast_i32_to_u32[65536]	1.8 ms	2.8 ms	-35.51%
⚡	Simulation	decompress[u64, (1000, 16)]	133.4 µs	65 µs	×2.1
⚡	Simulation	polygons	211.9 µs	167.2 µs	+26.74%
⚡	Simulation	linestrings	211.3 µs	167.1 µs	+26.49%
⚡	Simulation	points	208.2 µs	164.7 µs	+26.4%
⚡	Simulation	list_sum_small	280.4 µs	237.2 µs	+18.21%
⚡	Simulation	multipolygon_random_nulls	941.5 µs	811.9 µs	+15.97%
⚡	Simulation	nullable_points	396.6 µs	344.7 µs	+15.05%
⚡	Simulation	envelope_of_collected_points	559.4 µs	487.7 µs	+14.7%
⚡	Simulation	multipolygon_mixed_validity	1.2 ms	1.1 ms	+12.81%
⚡	Simulation	take_filter_list_slice_mask_sequential_indices[768, 50]	237.7 µs	211.2 µs	+12.55%
⚡	Simulation	multipolygon_non_nullable	1.2 ms	1.1 ms	+12.38%
⚡	Simulation	take_filter_list_random_mask_sequential_indices[768, 50]	233.4 µs	207.8 µs	+12.33%
⚡	Simulation	take_filter_list_small_random_mask_random_indices[256, 10]	232.5 µs	208.3 µs	+11.61%
⚡	Simulation	take_filter_list_slice_mask_sequential_indices[256, 50]	237.5 µs	213.1 µs	+11.47%
⚡	Simulation	take_filter_list_small_uncached_random_mask_random_indices[256, 10]	232.8 µs	209 µs	+11.42%
⚡	Simulation	take_filter_list_random_mask_sequential_indices[256, 50]	238.4 µs	214.1 µs	+11.34%
⚡	Simulation	take_filter_list_slice_mask_random_indices[256, 50]	236.1 µs	212.4 µs	+11.14%
⚡	Simulation	take_filter_list_nullable_random_mask_random_indices[256, 50]	256.8 µs	231.2 µs	+11.07%
⚡	Simulation	take_filter_list_small_uncached_random_mask_random_indices[768, 10]	240.6 µs	216.6 µs	+11.06%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

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

---

_{Comparing nemo/geo-convex-hull (4045fa5) with nemo/geo-collect (1f9f067)²}

Footnotes

1. 85 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on nemo/geo-collect (de907e5) during the generation of this report, so 64e3381 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[connortsui20]

LGTM