[common] Fix BinaryRowSerializer reuse buffer never shrinking

[yugan95]

Purpose

Linked issue: #7620

During external merge sort, each merge channel holds a BinaryRow reuse instance via BinaryRowSerializer.deserialize(reuse, source). When a large record is deserialized, the backing MemorySegment grows to fit it but is never shrunk for subsequent small records. With max-num-file-handles (default 128) channels each retaining a 100MB+ buffer, memory usage explodes into OOM.

Changes

• BinaryRowSerializer: add shrink logic with combined cap and ratio hysteresis in deserialize(BinaryRow reuse, DataInputView source) — reallocate only when both conditions are met:

  bufferSize > MAX_RETAINED_REUSE_BUFFER_SIZE (4MB)
      && bufferSize > recordLength * SHRINK_RATIO (4)

• 100MB buffer + 5MB record → ratio 20x > 4x → shrink
• 10MB buffer + 5MB record → ratio 2x < 4x → retain (no thrashing)
• 3MB buffer + tiny record → below 4MB cap → retain

The ratio-based approach avoids thrashing for sustained medium-to-large records while still reclaiming memory after a spike. Consistent with the approach used in #8159.

Tests

BinaryRowSerializerShrinkTest — 6 test cases covering:

• Spike followed by small record → shrink
• Spike followed by medium record → shrink (ratio check)
• Buffer proportional to record size → retain (ratio < 4x, even if > 4MB)
• Small buffer → retain
• Buffer grows when needed
• Consecutive large records → retain (no thrashing)

API and Format

N/A

Documentation

N/A

[yugan95]

Updated shrink logic to use a combined fixed-cap and ratio check, consistent with #8159:

  bufferSize > 4MB && bufferSize > recordLength * 4

This replaces the previous fixed-threshold-only approach (bufferSize > 4MB && length < 4MB) which would miss cases like a 100MB buffer with 5MB records (ratio 20x, clearly worth shrinking but previously retained because 5MB > 4MB).

Tests updated to 6 cases covering the ratio-based scenarios.

[JingsongLi]

Can you reuse fields in #8159

[yugan95]

Done. Reused RowHelper.MAX_RETAINED_REUSE_BUFFER_SIZE and RowHelper.SHRINK_RATIO from #8159, removed the duplicate definitions in BinaryRowSerializer.

[JingsongLi]

+1