[CALCITE-7343] RelToSqlConverter generate wrong sql when scalar correlated sub-query in Project

[Dwrite]

Summary

Fixes scalar correlated subquery SQL generation by ensuring proper table alias handling for correlation variables, particularly in deeply nested scenarios.

Changes

Core Strategy: Context-Aware Aliasing

• Always generates explicit aliases for correlated subqueries
• Prioritizes natural table identifiers (e.g., "EMP") over synthetic ones (e.g., "t") when the relation is a simple TableScan
• Uses correlation scope stack to support multi-level nested subqueries

Key Benefits:

1. DML Compatibility: Ensures UPDATE/DELETE statements remain valid in dialects that prohibit target table aliasing
2. SQL Readability: Produces intuitive SQL where correlations reference source table names
3. Multi-level Nesting: Correctly handles deeply nested correlated subqueries

Test Coverage

• ✅ Single-level scalar correlated subquery
• ✅ Deeply nested scalar correlated subqueries (2+ levels)
• ✅ UPDATE with correlated subquery (no unnecessary alias)
• ✅ Backward compatibility with existing correlation scenarios

Future Work

Follow-up PRs will extend this fix to:

• ⏳ Filter nodes: Correlated subqueries in WHERE/HAVING clauses
• ⏳ Join nodes: Correlated subqueries in JOIN conditions
• ⏳ Other correlation scenarios (EXISTS, IN, etc.)

The infrastructure (correlation scope stack, Result.resetAliasForCorrelation) is designed to be reusable for these future enhancements.

[xiedeyantu]

Please keep the PR title, commit message, and Jira title consistent with each other. Do not use "Fix" in the title.

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
87.9% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[xiedeyantu]

I have no further comments.

[mihaibudiu]

If there are no further comments I plan to merge this