Add GeoArrow test files for comprehensive parsing tests

[repro-code]

Summary

• Add test/links.txt with 127 small parquet files (<2MB) from geoarrow.org/data.html
• Update runtests.jl to download geoarrow test files to data/geoarrow/
• Add testset to verify GeoParquet.read() can parse all files, check :geometry column exists, and run GI.testgeometry on first geometry
• Change GeoParquet.read to warn instead of error when GeoParquet metadata is missing, returning the DataFrame without geometry parsing
• Add using WellKnownGeometry in tests to enable proper WKB geometry handling

Test files included

Files cover various geometry types and encodings:

• Natural Earth (12 files): cities, countries, countries-geography, countries-bounds
• Quadrangles (3 files): 100k quadrangles
• CRS Examples (11 files): Vermont with different CRS encodings
• Geometry Examples (98 files): Point, LineString, Polygon, MultiPoint, MultiLineString, MultiPolygon, Geometry, GeometryCollection with Z/M/ZM variants
• NS Water (3 files): water-point data (~1.7-1.9MB)

Each dataset has _geo (GeoArrow encoding) and _native (native Arrow encoding) variants where applicable.

Changes to GeoParquet.read behavior

Files without valid GeoParquet metadata now emit a warning and return the DataFrame as-is, rather than throwing an error. This allows reading plain parquet files that happen to have geometry columns but no geo metadata.

Test plan

• Verify all 127 files download successfully
• Check which files fail to parse and categorize errors
• Document any parsing issues for future fixes

🤖 Generated with Claude Code