fix groupconcat

[daviszhen]

What type of PR is this?

• API-change
• BUG
• Improvement
• Documentation
• Feature
• Test and CI
• Code Refactoring

Which issue(s) this PR fixes:

issue #24359

What this PR does / why we need it:

修改 group_concat 的order by 改写.

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[fengttt]

I am very surprised we need such a function. This is generic order by clause and should reuse all the code from say, sort operator.

More: type is ALWAYS available, so we should NOT need a type switch.

[XuPeng-SH]

Requesting changes for two semantic regressions in the new GROUP_CONCAT ... ORDER BY rewrite.

1. DISTINCT + ORDER BY can keep the wrong representative row

   • In concat2.go, DISTINCT dedup happens first via distinctHash.fill(...) on only the concat arguments.
   • The ORDER BY payload is sorted later in flushGroup().
   • So for cases like GROUP_CONCAT(DISTINCT x ORDER BY y), duplicate x values with different y values now keep whichever row was seen first, not the one implied by the ORDER BY.

2. Explicit NULLS FIRST / NULLS LAST is dropped

   • bindGroupConcatOrderBy() in having_binder.go only records ASC/DESC into orderFlags.
   • Execution then compares nils with the default behavior in types.CompareValue, so explicit null placement is no longer preserved.

I think both need to be fixed or rejected explicitly before approval.