Parquet: don't truncate min/max statistics for float16 and decimal when writing file

[Jefffrey]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

See discussion:

#5003 (comment)

#5003 added support for float16 type which is a logical type on top of fixed len byte array

When writing statistics, truncation can occur for binary physical type:

arrow-rs/parquet/src/column/writer/mod.rs

Lines 634 to 643 in 7ba36b0

	// We only truncate if the data is represented as binary
	match self.descr.physical_type() {
	Type::BYTE_ARRAY | Type::FIXED_LEN_BYTE_ARRAY => {
	self.column_index_builder.append(
	null_page,
	self.truncate_min_value(stat.min_bytes()),
	self.truncate_max_value(stat.max_bytes()),
	self.page_metrics.num_page_nulls as i64,
	);
	}

Which might be troublesome for f16 type, if the column_index_truncate_length config is set to 1, as a truncated f16 wouldn't represent the min and max correctly anymore as it has a sort order different from fixed len byte array

Describe the solution you'd like

Ignore truncation for f16 when writing min/max statistics

Describe alternatives you've considered

Additional context

Do we need to worry about this for other types based on binary physical types? i.e. decimal

[Jefffrey]

Potentially has relation to #5037?

[tustvold]

label_issue.py automatically added labels {'parquet'} from #5154