fix(transforms): day transform result type is date, not int

[twuebi]

DayTransform.ResultType returned Int32, diverging from Java, Rust, and PyIceberg, which all return date. The physical value stays int32 days, the date type adds the date Avro logicalType to the partition field, so day partitions round-trip with the same logical type Spark writes.

Without it, compacting a days partitioned table whose mainfest was written by spark failed on commit:

avro: field data_file.partition.ts_day: cannot use time.Time with Avro type intavro: field data_file.partition.ts_day: cannot use time.Time with Avro type int (*fmt.wrapError)

[laskoviymishka]

[laskoviymishka]

@twuebi see: https://lists.apache.org/thread/wnhvdmwc3b1dxc2ogk7po682pky12cps and #915

[laskoviymishka]

One more point to connect dotes - apache/iceberg#16446

[twuebi]

oh, sorry for not checking issues / open PRs first, this blew up compaction running on a table with partition by day written by spark