perf: investigate the performance of DataChunkBuilder

[lmatz]

Proposed by @BugenZhao @st1page

[kwannoel]

Any specific directions or existing evidence that this can be further optimized?

• For instance does it show up on q17 flamegraph?
• Or if a benchmark doesn't currently exist, but we suspect DataChunkBuilder might not be that performant, what should be the spec of the benchmark?
• Or some other observation?

[lmatz]

The motivation comes from #9300 instead of observation.

If building a new chunk is slow, then the optimization discussed in #9300 leads to unfair comparison and thus we could make a completely different conclusion regarding whether #9300 improves performance or not.

[github-actions]

This issue has been open for 60 days with no activity. Could you please update the status? Feel free to continue discussion or close as not planned.