This PR adds a basic Parquet file reading capability to DuckDB. For now, this reader does not support nested tables. Some higher-level types are also still unsupported due to lack of agreement between systems. For now we only support Snappy compression.

Use like so from SQL:

SELECT * FROM PARQUET_SCAN('some/parquet/file.parquet')

The parquet reader is also integrated with the Python relational API, e.g.

import duckdb
duckdb.from_parquet('some/parquet/file.parquet').limit(10)

The Parquet reader is implemented as a DuckDB extension. The Python and R packages load this extension by default, from C++ :

#include "parquet-extension.hpp"
db.LoadExtension<ParquetExtension>()