Optimize Fast Compaction performance

In the scenario where all seq pages overlap with unseq pages (for example, the sequence point is 2 4 8, and the unsequence point is 1 3 5 7, etc), with four sub-threads, Fast performance is slower than ReadPoint.