Docs: add recommendations about parquet writing options

For both LSDB and hats-import we should provide some tips on `pyarrow.parquet.write_table` option optimization.

For example, I have learned recently that for list-columns with repeated lists enabling `use_byte_stream_split=[list_column], use_dictionary=[col for col in columns if col != list_column], compression='ZSTD', compression_level=15` leads to incredibly efficient compression. I've tested it on [this file](https://huggingface.co/datasets/MultimodalUniverse/desi/blob/main/edr_sv3/train-00000-of-00007.parquet) and have improved the compression ratio from ×2.5 to ×66 for `spectrum.lambda` column, first chunk.

Locate the documentation for both LSDB and hats-import and identify where parquet-writing guidance belongs. Start by reviewing the pyarrow.parquet.write_table options and the provided compression example, then document applicable recommendations for list-columns and other columns in both projects. Done means contributors can find and apply the tips when writing parquet files.

Written by the indexing model from the issue text.