Track memory used by parquet writers. #11344

Is your feature request related to a problem or challenge?

The encoding of parquet requires a non-trivial amount of memory buffering. During the execution of datafusion physical plans, parquet may be encoded using ParquetSink (e.g. COPY TO queries which output parquet). Currently we do not track ParquetSink's memory usage in the task context's memory pool.

Describe the solution you'd like

Start tracking the memory used during parquet encoding.

Describe alternatives you've considered

No response

Additional context

Recently, we extended several parquet interfaces to provide better estimates of the memory_usage during encoding. These memory usage estimates should be used to determine the appropriate memory reservations.