Support reading INT96 timestamps as LongTimestampWithTimeZoneType

[raunaqmorarka]

Hive tables migrated to iceberg through Spark require the ability to read INT96 timestamps as LongTimestampWithTimeZoneType

Description

Fixes #11338

Additional context and related issues

Release notes

( ) This is not user-visible or is docs only, and no release notes are required.
( ) Release notes are required. Please propose a release note for me.
(x) Release notes are required, with the following suggested text:

# Iceberg
* Fix failure to read hive tables migrated to iceberg through Apache Spark. ({issue}`11338`)

[wendigo]

Ship it!

[findinpath]

Pls add in the description of the PR that the following method can be used as reference while reviewing the PR

trino/lib/trino-parquet/src/main/java/io/trino/parquet/reader/decoders/ValueDecoders.java

Lines 407 to 434 in eda8fdf

	public ValueDecoder<int[]> getInt96ToLongTimestampDecoder(ParquetEncoding encoding, DateTimeZone timeZone)
	{
	checkArgument(
	field.getType() instanceof TimestampType timestampType && !timestampType.isShort(),
	"Trino type %s is not a long timestamp",
	field.getType());
	int precision = ((TimestampType) field.getType()).getPrecision();
	return new InlineTransformDecoder<>(
	getInt96TimestampDecoder(encoding),
	(values, offset, length) -> {
	for (int i = offset; i < offset + length; i++) {
	long epochSeconds = decodeFixed12First(values, i);
	long nanosOfSecond = decodeFixed12Second(values, i);
	if (timeZone != DateTimeZone.UTC) {
	epochSeconds = timeZone.convertUTCToLocal(epochSeconds * MILLISECONDS_PER_SECOND) / MILLISECONDS_PER_SECOND;
	}
	if (precision < 9) {
	nanosOfSecond = (int) round(nanosOfSecond, 9 - precision);
	}
	// epochMicros
	encodeFixed12(
	epochSeconds * MICROSECONDS_PER_SECOND + (nanosOfSecond / NANOSECONDS_PER_MICROSECOND),
	(int) ((nanosOfSecond * PICOSECONDS_PER_NANOSECOND) % PICOSECONDS_PER_MICROSECOND),
	values,
	i);
	}
	});
	}

[findinpath]

Is this code section tested anywhere?

have .123456789 in the storage , but having MICROSECONDS or MILLISECONDS the precision being read?

[raunaqmorarka]

Yes, this is called when TIMESTAMP_TZ_MICROS is tested, io.trino.parquet.reader.TestingColumnReader#WRITE_INT96 populates nanos which result in rounding to lower precision

[wendigo]