[FEA] Use pyarrow to read and serialize geoPandas geometry

[thomcom]

Is your feature request related to a problem? Please describe.
In #575, I'm correcting an error with my GeoArrow implementation. The specification in pyarrow for polygons and linestrings is

polygon_type = pa.list_(
    pa.field("polygons", pa.list_(
        pa.field("rings", pa.list_(
            pa.field("vertices", pa.list_(
                pa.field("xy", pa.float64(), nullable=False), 2),
            nullable=False)), nullable=False))))

linestring_type = pa.list_(
    pa.field("lines", pa.list_(
        pa.field("offsets", pa.list_(
            pa.field("xy", pa.float64(), nullable=False), 2),
        nullable=False)), nullable=False))

Because I implemented arrow from scratch in #300, and in a spasm of reasoning, I forgot to use the prefix-scan method for indexing the above types.

The first step in correcting it, given where I am with pyarrow now, is to change the input of our GeoArrow types to use arrow directly.

Describe the solution you'd like

cuspatial/io/geopandas_adapter.py implements the arrow spec when reading from GeoSeries objects in a GeoPandas dataframe.

_load_geometry_offsets computes the sizes and prefix sum/scan buffers for each of the four arrow lists and allocates memory for them.

_read_geometries iterates over the geometries again, copying individual coordinates and offsets into the correct position in each buffer.

Instead, specify the GeoArrow format types as per above in geoarrow.py, then use those formats and densely packed tuple buffers from GeoSeries[i].__geo_interface__['coordinates'] to create the geoarrowbuffers.py objects.

This will break unit tests for LineStrings and MultiLineStrings, which have been implemented incorrectly, leading to the original Issue #575 which will have another PR to fix them.