Integrate duckDB queries into worldcereal-classification methodology

[kvantricht]

@cbutsko please mention here the code snippet we want to integrate.

[cbutsko]

def get_spatial_subset(bbox_poly):
    db = duckdb.connect()
    # db.sql('INSTALL spatial')
    db.load_extension('spatial')

    parquet_fpath = 's3://geoparquet/worldcereal_extractions_phase1/*/*.parquet'

    query_df = db.sql(f"""
    set s3_endpoint='s3.waw3-1.cloudferro.com';
    select *
    from read_parquet('{parquet_fpath}', hive_partitioning = 1) original_data
    where st_within(ST_Point(original_data.lon, original_data.lat), ST_GeomFromText('{bbox_poly.wkt}'))
    """).df()
    return query_df