[WIP][SPARK-50361][SQL] Fix bugs of using Lateral Column Alias in ORDER BY

[yaooqinn]

What changes were proposed in this pull request?

1. Aggreges using Lateral Column Alias in ORDER BY should throw LATERAL_COLUMN_ALIAS_IN_AGGREGATE_FUNC
2. AMBIGUOUS_LATERAL_COLUMN_ALIAS if there are 2 or more matches of lateral column aliases

Why are the changes needed?

For 1,
An invalid case, such as SELECT MAX(id) AS id2 FROM range(1) ORDER BY AVG(id2), is considered valid in versions before 3.2.x. Since version 3.2, a refactoring pull request accidentally 'fixes' this case; however, the error message is not straightforward.

[MISSING_ATTRIBUTES.RESOLVED_ATTRIBUTE_MISSING_FROM_INPUT] Resolved attribute(s) "id2" missing from "id" in operator !Aggregate [max(id#130L) AS id2#129L, avg(id2#129L) AS avg(id2#129L)#133].  SQLSTATE: XX000; line 1 pos 36;
Project [id2#129L]
+- Sort [avg(id2#129L)#133 ASC NULLS FIRST], true
   +- !Aggregate [max(id#130L) AS id2#129L, avg(id2#129L) AS avg(id2#129L)#133]
      +- Range (0, 1, step=1)

This error message somewhat confuses users when debugging/upgrading.

For 2, it's bugfix, we shall not ignore the AMBIGUOUS_LATERAL_COLUMN_ALIAS but go further resolution

In this pull request, we explicitly address this issue as we did for standard aggregates.

Does this PR introduce any user-facing change?

Yes, Aggregates using Lateral Column Alias in ORDER BY throws LATERAL_COLUMN_ALIAS_IN_AGG instead of RESOLVED_ATTRIBUTE_MISSING_FROM_INPUT.

How was this patch tested?

new tests

Was this patch authored or co-authored using generative AI tooling?

NO