[Binlog][Cloud] Compaction commit response loss misclassifies durable success as failure

## Problem

[Binlog][Cloud] Compaction commit response loss misclassifies durable success as failure

## Minimal reproduction path

In a Cloud ROW Binlog workload, force the response of a compaction commit to be lost after the commit is durable, then retry the same operation. Verify that the retry recognizes the durable commit, does not report a false failure, and does not duplicate or discard row-binlog data.

## Expected result

A durable compaction commit remains successful across response loss and retry, without duplicate data.

## Source

This report is a sanitized tracking record for Jira DORIS-28070. Internal hosts, paths, artifacts, build identifiers, and private logs are intentionally omitted.

No source files, tests, or entry points are identified; begin by locating the Cloud ROW Binlog compaction commit and retry handling, then reproduce response loss after a durable commit. Done means a retry recognizes the durable commit as successful, reports no false failure, and neither duplicates nor discards row-binlog data.

Written by the indexing model from the issue text.