fix: plan UNION of untyped aliased placeholders

[adriangb]

Which issue does this PR close?

• Closes Untyped aliased placeholder in a UNION arm fails with "Schema error: No field named <alias>" when the plan is analyzed or optimized #25669.

Rationale for this change

PREPARE p AS SELECT $1 AS a UNION ALL SELECT $2 AS a fails with Schema error: No field named a. The same happens when a caller optimizes such a plan before binding its parameters.

Expr::get_type had a special case for an alias of an untyped placeholder: it looked up the alias name as a column in the schema it receives. The UNION coercion code calls get_type on each projection expression with the projection's input schema, which does not contain the alias, so the lookup fails. Expr::to_field has no such special case and returns a Null field for the same expression.

The special case came from #4701. At that time get_type returned an error for a bare untyped placeholder. Today a bare untyped placeholder has type Null, so the special case is no longer needed.

What changes are included in this PR?

• Remove the Alias(Placeholder { field: None }) special case from Expr::get_type. An alias now has the type of the expression it wraps, the same as in to_field.

What is the testing strategy for this PR?

Tests added:

• sqllogictest cases in prepare.slt for PREPARE and EXECUTE of a UNION ALL of aliased placeholders, including a CTE joined to a table.
• A test in core/tests/sql/select.rs that optimizes the plan with unbound placeholders, then binds and executes it.
• A unit test in expr_schema.rs that get_type and to_field agree for an aliased untyped placeholder.

The existing prepared statement and placeholder type inference tests pass without changes.

Are there any user-facing changes?

The queries above now plan and run. No API changes.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.48%. Comparing base (95bb0a0) to head (fd981db).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #25675      +/-   ##
==========================================
- Coverage   82.48%   82.48%   -0.01%     
==========================================
  Files        1140     1140              
  Lines      437614   437622       +8     
  Branches   437614   437622       +8     
==========================================
+ Hits       360986   360987       +1     
- Misses      54834    54840       +6     
- Partials    21794    21795       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[adriangb]

@neilconway could I ping you for review?

[dustinCodes84600]

This is useful for a Rust service where we cache optimized DataFusion logical plans before binding request params. Hit the same schema lookup failure when a parameterized two-row key set used UNION ALL before joining an Arrow-backed table.

fwiw, while reading I made a visual walkthrough covering alias type resolution: https://flyovers.dev/flyover/85a1f93d-380b-43e5-9401-274f9f7e6e81

[adriangb]

This is useful for a Rust service where we cache optimized DataFusion logical plans before binding request params. Hit the same schema lookup failure when a parameterized two-row key set used UNION ALL before joining an Arrow-backed table.

fwiw, while reading I made a visual walkthrough covering alias type resolution: https://flyovers.dev/flyover/85a1f93d-380b-43e5-9401-274f9f7e6e81

I was not expecting what I got when I clicked that link 🤣. If you care about this PR maybe a GitHub review would be helpful in helping get it across the line? Or confirmation that it fixes your problem?