Trino doesn't parallelize OPTIMIZE within a partition

[itamarwe]

The Trino Iceberg connector was intentionally limited to a single writer per partition to avoid small files (#18388).
Nevertheless, it means that the optimization throughput for each partition is limited to the throughput of a single writer.
For use-cases with high write throughputs per partition, it's a limitation that might not allow performing optimization at all.

Ideally, we would scale the writers within a partition with regards to the expected number of files to be written. We know at the beginning of the optimization the volume of the data pending optimization in each partition and can estimate how many files will be written within each partition.

That will greatly accelerate optimization.