feat(r): Add GDAL read to R package

[paleolimbot]

In Python we can read GDAL/OGR sources with read_pyogrio():

sedona-db/python/sedonadb/python/sedonadb/datasource.py

Lines 112 to 172 in e80fb37

	class PyogrioFormatSpec(ExternalFormatSpec):
	"""An `ExternalFormatSpec` implementation wrapping GDAL/OGR via pyogrio"""
	def __init__(self, extension=""):
	self._extension = extension
	self._options = {}
	def with_options(self, options):
	cloned = type(self)(self.extension)
	cloned._options.update(options)
	return cloned
	@property
	def extension(self) -> str:
	return self._extension
	def open_reader(self, args):
	import pyogrio.raw
	url = args.src.to_url()
	if url is None:
	raise ValueError(f"Can't convert {args.src} to OGR-openable object")
	if url.startswith("http://") or url.startswith("https://"):
	ogr_src = f"/vsicurl/{url}"
	elif url.startswith("file://") and sys.platform != "win32":
	ogr_src = url.removeprefix("file://")
	elif url.startswith("file:///"):
	ogr_src = url.removeprefix("file:///")
	else:
	raise ValueError(f"Can't open {url} with OGR")
	if ogr_src.endswith(".zip"):
	ogr_src = f"/vsizip/{ogr_src}"
	if args.is_projected():
	file_columns = args.file_schema.names
	columns = [file_columns[i] for i in args.file_projection]
	else:
	columns = None
	batch_size = args.batch_size if args.batch_size is not None else 0
	if args.filter and args.file_schema is not None:
	geometry_column_indices = args.file_schema.geometry_column_indices
	file_columns = args.file_schema.names
	if len(geometry_column_indices) == 1:
	bbox = args.filter.bounding_box(
	file_columns[geometry_column_indices[0]]
	)
	else:
	bbox = None
	else:
	bbox = None
	return PyogrioReaderShelter(
	pyogrio.raw.ogr_open_arrow(
	ogr_src, {}, columns=columns, batch_size=batch_size, bbox=bbox
	),
	columns,
	)

This works by leveraging the ArrowArrayStream API as exposed by pyogrio so we don't need to ship a separate GDAL build.

In R it's a significant performance hit to have this read go materialize to sf ( apache/sedona#2576 ). We may need to add something to the sf package to allow the read to skip materialization; however, the sf package does implement the array stream interface at least a little: https://github.com/r-spatial/sf/blob/25700c2cb48191598bfc7495baafe4b6808398c6/R/read.R#L221-L265 , possibly enough to prototype.