DataFrame WriteAsync always creates nullable data types in schema

[thomasvdb]

Issue description

Is there a reason why it's defaulting to nullable data types when writing DataFrame back to a Stream?
See

parquet-dotnet/src/Parquet.Data.Analysis/Extensions.cs

Line 65 in 1846cc7

df.Columns.Select(col => new DataField(col.Name, col.DataType.GetNullable())));

var df = await parquetFileStream.ReadParquetAsDataFrameAsync(cancellationToken: cancellationToken);
var sortedDataFrame = df.OrderBy("ReadingValue");

var newStream = new MemoryStream();
await sortedDataFrame.WriteAsync(newStream, cancellationToken: cancellationToken);

The parquetFileStream variable contains a schema where we have a non-nullable DateTime.
Afterwards the newStream variable contains a schema where we have a nullable DateTime.

And this returns an InvalidDataException when deserializing the newStream variable again with ParquetSerializer.DeserializeAsync

System.IO.InvalidDataException: class definition level (0) does not match file's definition level (1) in field 'ReadingDateTime'. This usually means nullability in class definiton is incompatible.

[thomasvdb]

Looked furter into it and it seems (correct me if I'm wrong) that DataFrame doesn't keep track if the type was nullable when reading the value.