Support RLE, DELTA_BYTE_ARRAY and DELTA_LENGTH_BYTE_ARRAY Parquet encodings

[Mytherin]

Fixes #3981

This should implement all remaining Parquet encoding types - aside from Byte Stream Split which does not appear to be used by anyone anywhere.

[dselivanov]

@Mytherin what is the way to try it in python? Should I follow build steps from "python" github actions or there is a simpler way?

[Mause]

@Mytherin what is the way to try it in python? Should I follow build steps from "python" github actions or there is a simpler way?

You can install the Python prerelease using the pip --pre flag

[dselivanov]

Thanks @Mause. I believe --pre installs from master branch (and current feature is still in feature branch)? I've tried pip install duckdb --pre --upgrade and see Unsupported page encoding while reading parquet (duckdb=0.6.1.dev86)

[dselivanov]

I was able to build package on linux machine just by running python3 -m build from within ./tools/pythonpkg. Reading parquet files with DELTA_LENGTH_BYTE_ARRAY encoded columns works smooth now.
On my local OSX machine, however, package built from feature branch throws segmentation fault when running import duckdb. Mb something with my environment - not sure.

[Mytherin]

That sounds like something funky with your environment indeed, perhaps try uninstalling all versions of duckdb (repeat pip uninstall duckdb until no more versions remain) and doing a clean build?