Version 0.5.0

Major enhancements

• Support for reading based on Arrow as the transfer mechanism of the data
  from GDAL to Python (requires GDAL >= 3.6 and pyarrow to be installed).
  This can be enabled by passing use_arrow=True to pyogrio.read_dataframe
  (or by using pyogrio.raw.read_arrow directly), and provides a further
  speed-up (#155, #191).
• Support for appending to an existing data source when supported by GDAL by
  passing append=True to pyogrio.write_dataframe (#197).

Potentially breaking changes

• In floating point columns, NaN values are now by default written as "null"
  instead of NaN, but with an option to control this (pass nan_as_null=False
  to keep the previous behaviour) (#190).

Improvements

• It is now possible to pass GDAL's dataset creation options in addition
  to layer creation options in pyogrio.write_dataframe (#189).
• When specifying a subset of columns to read, unnecessary IO or parsing
  is now avoided (#195).

Packaging

• The GDAL library included in the wheels is updated from 3.4 to GDAL 3.6.2,
  and is now built with GEOS and sqlite with rtree support enabled
  (which allows writing a spatial index for GeoPackage).
• Wheels are now available for Python 3.11.
• Wheels are now available for MacOS arm64.