feat(stream): support non-append-only process time temporal join

[chenzl25]

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

What's changed and what's your intention?

• Resolve Feat: support non-append-only process time temporal join  #15218
• Support non-append-only process time temporal join which will introduce a state called memo_table to the temporal join.
• memo_table's pk is join_key + left_pk + right_pk. And its columns are almost the same as the right table.
• Write pattern:
  for each left input row (with insert op), construct the memo table pk and insert the row.
  Read pattern:
  for each left input row (with delete op), construct pk prefix (join_key + left_pk) to fetch rows.

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• I have added test labels as necessary. See details.
• I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
• My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
• All checks passed in ./risedev check (or alias, ./risedev c)
• My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

• My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

• Support non-append-only temporal join, that is the outer side of the temporal join doesn't require to be append-only anymore.

[BugenZhao]

This looks good.

Correct me if I'm wrong: we face the exactly same issue with indeterministic UDFs. Is it possible to make the memoization(materialization) into a separate operator? So that we can reuse it for these similar cases.

[TennyZhuang]

we face the exactly same issue with indeterministic UDFs. Is it possible to make the memoization(materialization) into a separate operator

There should be an abstraction for the logic, but IMO it shouldn’t be a separate operator.

[chenzl25]

This looks good.

Correct me if I'm wrong: we face the exactly same issue with indeterministic UDFs. Is it possible to make the memoization(materialization) into a separate operator? So that we can reuse it for these similar cases.

Even though indeterministic UDFs also need a memo table, I don't think we can use a shared operator to handle both cases. Because the memo table access pattern and the way how to reconstruct rows from the memo table are totally different.

[BugenZhao]

Because the memo table access pattern and the way how to reconstruct rows from the memo table are totally different.

You're right. I was solely focused on using temporal join for point-lookup. 😄

[st1page]

LGTM, maybe we can notice user their will be a memo table state for non-append-only temporal join to prevent silently generating a stream plan with bad performance.(One case I can come up with is #13549)

[fuyufjh]

I think we need to handle the case that a row is updated while the join key is not changed. This can be done in a separate PR, but overall I think it's a must-have enhancement