fix(stream): avoid committing to in-memory state store in join benchmark

[kwannoel]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

If we commit to state store, we will end up measuring the memory consumption of writing the amplified chunks to the in-memory state store.
In production, state store is on top of object storage / disk. The in-memory state store is only used in benchmark / tests. So we shouldn't benchmark the memory consumption.

Further, the amplified chunks are not written in a large batch typically. The scenario we care about is OOM from dimension table update. In such a scenario, the fact table side's memory consumption will be high due to reads from the update side, rather than writes.

Checklist

• I have written necessary rustdoc comments.
• I have added necessary unit tests and integration tests.
• I have added test labels as necessary.
• I have added fuzzing tests or opened an issue to track them.
• My PR contains breaking changes.
• My PR changes performance-critical code, so I will run (micro) benchmarks and present the results.
• My PR contains critical fixes that are necessary to be merged into the latest release.

Documentation

• My PR needs documentation updates.

Release note

[kwannoel]

• perf(stream): concurrently fetch row from storage and refill cache #19629 : 2 dependent PRs (#20025 , #20041 )
• fix(stream): avoid committing to in-memory state store in join benchmark #20008 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[kwannoel]

This second barrier will trigger a state store flush. Because the barrier before this is the initial barrier, it will not commit data to state store.