Describe the enhancement requested

We should improve the set of data types in our Parquet fuzz seeding corpus.
Any Arrow data type that can trigger a specific decoding/conversion sequence can apply.
Examples: Boolean, FixedSizeBinary, Decimals, Temporals...

This can be paired with a more recent Parquet format version, and storing the Arrow schema to ensure the original data type is decoded on read.

Component(s)

C++, Parquet