[parquet] Support us precision Time type in batch reader

[ZacBlanco]

Description

Iceberg timestamps are written with microsecond precision as defined by the spec. Before this change, the parquet batch reader optimization did not properly support reading microsecond-precision time types. Presto returned results which differed depending on whether the optimization was enabled

This change adds proper support in the batch reader.

Motivation and Context

Previously Presto would return wrong results in the Iceberg connector when parquet batch read optimizations are enabled.

Behavior before this change:

CREATE TABLE t(i time);
INSERT INTO t VALUES time '1:00';
-- without batch optimization
presto:tpch> SET SESSION iceberg.parquet_batch_read_optimization_enabled = false;
presto:tpch> SELECT CAST(i as bigint) FROM t;
  _col0
----------
 32400000 -- (5 trailing zeroes)
(1 row)


-- with batch optimization
presto:tpch> SET SESSION iceberg.parquet_batch_read_optimization_enabled = true;
SET SESSION
presto:tpch> SELECT CAST(i as bigint) FROM t;
    _col0
-------------
 32400000000 -- (8 trailing zeroes)
(1 row)

Behavior after this change:

CREATE TABLE t(i time);
INSERT INTO t VALUES time '1:00';
-- without batch optimization
presto:tpch> SET SESSION iceberg.parquet_batch_read_optimization_enabled = false;
presto:tpch> SELECT CAST(i as bigint) FROM t;
  _col0
----------
 32400000 -- (5 trailing zeroes)
(1 row)


-- with batch optimization
presto:tpch> SET SESSION iceberg.parquet_batch_read_optimization_enabled = true;
SET SESSION
presto:tpch> SELECT CAST(i as bigint) FROM t;
    _col0
-------------
 32400000 -- (5 trailing zeroes)
(1 row)

Impact

Proper time type support.

Test Plan

Added test to verify equivalent return values with and without the optimization in the Iceberg connector.

Contributor checklist

• Please make sure your submission complies with our development, formatting, commit message, and attribution guidelines.
• PR description addresses the issue accurately and concisely. If the change is non-trivial, a GitHub Issue is referenced.
• Documented new properties (with its default value), SQL syntax, functions, or other functionality.
• If release notes are required, they follow the release notes guidelines.
• Adequate tests were added if applicable.
• CI passed.

Release Notes

== RELEASE NOTES ==

Iceberg Connector Changes
* Fix time-type columns to return properly when ``iceberg.parquet-batch-read-optimization-enabled`` is set to ``TRUE``. :pr:`23542`

[steveburnett]

Nit suggestion for release note entry:

== RELEASE NOTES ==

Iceberg Connector Changes
* Fix time-type columns to return properly when ``iceberg.parquet-batch-read-optimization-enabled`` is set to ``TRUE``. :pr:`23542`

[hantangwangd]

Thanks for the fix. Lgtm, one little nit.

[yingsu00]

@ZacBlanco I see the Time decoders are exactly the same with existing Timestamp ones. There's no point to duplicate so much code. Maybe we can rename the existing Timestamp decoders like Int64TimestampMicrosRLEDictionaryValuesDecoder to Int64TimeAndTimestampMicrosRLEDictionaryValuesDecoder?

[ZacBlanco]

Thanks for the review @yingsu00 . I agree - let's try not to duplicate code. I updated the naming of the decoders across the module and updated the conditionals to check isTimestampMicros || isTimeMicros