Track uncompressed data size for varchar and varbinary in Iceberg

[findepi]

Iceberg column_sizes is

Map from column id to the total size on disk of all regions that store the column. Does not include bytes necessary to read other columns, like footers.

(https://iceberg.apache.org/spec/)

We use this field to fill io.trino.spi.statistics.ColumnStatistics#dataSize in Iceberg, but this should be uncompressed data size.

• [x] Short-term fix would be to heuristically account for compression (e.g. assume 10x compression). (see @danielcweeks 's https://trinodb.slack.com/archives/CJ6UC075E/p1669057878701609?thread_ts=1669024481.507009&cid=CJ6UC075E)
• [ ] Long-term
  • either we should add another field to Iceberg file metadata, that would be set for variable-width columns only (for fixed-width types in-memory size may be engine-dependent, see @danielcweeks 's https://trinodb.slack.com/archives/CJ6UC075E/p1669049890567839?thread_ts=1669024481.507009&cid=CJ6UC075E)
    • initially we should do this for varchar, varbinary columns
  • or, during ANALYZE collect varchar/varbinary length and store it in stats file
• [ ] (somewhat related) return L as ColumnStatistics#dataSize for fixed(L) columns

[findepi]

cc @danielcweeks @homar @ebyhr @alexjo2144

[homar]

@findepi I think short term is done (https://github.com/trinodb/trino/pull/15186)