Compress in-memory shuffled record batches

[richox]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

currently, shuffle writer keeps record batches uncompressed until they are spilled or flushed to next stage. if we compress them before spilling, we can reduce memory usage and the amount of spilled data.

Describe the solution you'd like

when a record batch is finished, instead of buffering it into a Vec<RecordBatch>, we compress it into aVec[u8] which is later spilled or flushed.

Describe alternatives you've considered

Additional context