Can we replace ParquetSharp.DataFrame with use of To/FromArrowRecordBatch

[adamreeve]

We made ParquetSharp.DataFrame for reading and writing to/from .NET DataFrames, but now that we have an Arrow based API for ParquetSharp it might make more sense to use DataFrame.FromArrowRecordBatch and DataFrame.ToArrowRecordBatches. In theory this should be higher performance by avoiding extra copies currently required by ParquetSharp.DataFrame.

This issue covers adding documentation to ParquetSharp on how to read and write DataFrames via Arrow and investigating whether there's any reason why users might still want to use ParquetSharp.DataFrame rather than Arrow.