[mihaibudiu]

Jira Link

CALCITE-7933

Changes Proposed

This commit contains multiple small changes and fixes related to lambdas:

SqlToRelConverter:

• Allow a lambda in a sub-query to read a column of the enclosing query.
• A lambda in an aggregate gave wrong results or an out-of-bounds error; now it should work correctly.

CorrelateProjectExtractor:

• does not move an expression that reads a parameter out of the lambda (the actual issue subject)

EnumUtils:

• In the generated Java code the parameters of the generated code are final. Before, Janino rejected the generated code.

Validator:

• Reject lambda which have subquries in the body. This is in line with all major SQL dialects which support lambdas.
• Reject lambdas which invoke an aggregate function, GROUPING, or a windowed aggregate.
• Reject a lambda in the ORDER BY clause which attempts to read columns (this used to crash; support can be added later if desired)
• SqlLambdaScope.isParameter checks the first name of a compound identifier, such as r.f0. Before, it failed with an AssertionError.

[mihaibudiu]

I have validated all quidem queries with DuckDB and Spark (Spark rejects two queries)