feat(vortex-geo): native Geometry extension type with point support

[HarukiMoriarty]

Summary

Alternative approach to #8372. That PR gives each geometry kind its own extension type (vortex.geo.point, …); this PR uses a single type vortex.geo.geometry for all kinds, with the kind and the CRS carried in the extension metadata.

Spatial functions are signed distance(geometry, geometry), not per-kind: the operand check is one ext.is::<Geometry>(), kernels dispatch on kind at runtime, and a new kind is a metadata value plus a match arm — not a new type and new signatures. The kind lives in metadata because GeoArrow layouts collide (linestring ≡ multipoint, polygon ≡ multilinestring).

Like #8372, point is the first supported kind.

API Changes

Adds to vortex-geo, all registered through vortex_geo::initialize:

• Extension type Geometry (vortex.geo.geometry).
• GeoMetadata gains a geometry_type: GeometryKind field next to crs.
• GeometryValue: the value a geometry scalar unpacks to; its Display emits WKT.
• Scalar function GeoDistance (vortex.geo.distance): per-row distance with PostGIS ST_Distance semantics.

Testing

Unit tests cover dtype validation for every GeoArrow dimension (plus rejection of invalid storage, non-point kinds, and the Unspecified kind), metadata round-tripping with and without a kind, round-tripping a point column through scalar execution back to the original coordinates, WKT display for every kind including dimension tags and EMPTY, distance over all operand shapes (column-to-constant on either side, column-to-column, constant-to-constant), and construction-time rejection of non-geometry operands, mismatched CRS, and nullable columns.

[codspeed-hq]

Merging this PR will improve performance by 26.14%

⚠️ 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.

⚡ 5 improved benchmarks
✅ 1531 untouched benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	chunked_bool_canonical_into[(1000, 10)]	35.5 µs	20.5 µs	+72.69%
⚡	Simulation	chunked_varbinview_canonical_into[(1000, 10)]	198.1 µs	162 µs	+22.32%
⚡	Simulation	chunked_varbinview_into_canonical[(1000, 10)]	213.5 µs	177.4 µs	+20.4%
⚡	Simulation	bitwise_not_vortex_buffer_mut[128]	244.4 ns	215.3 ns	+13.55%
⚡	Simulation	bitwise_not_vortex_buffer_mut[1024]	304.7 ns	275.6 ns	+10.58%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing nemo/geo-geometry (ee1e166) with develop (8acef3a)}