Support configuring compression levels for ZSTD writes

[akki]

Various Trino connectors support writing data with ZSTD compression (for e.g. using hive.compression-codec for Hive connector).
ZSTD also supports compression level (currently from 1 to 22). It would be good to support this functionality in Trino too, so that we can benefit from higher compression levels (reducing storage costs).

This is similar to what Spark allows with spark.sql.parquet.compression.coded.zstd.level.

[ebyhr]

cc: @raunaqmorarka