Handle relation metadata column names case-insensitively

[tom-ph]

Summary

Resolve the metadata columns returned by list_relations_without_caching case-insensitively.

Some Netezza environments or drivers preserve unquoted column aliases as lowercase or mixed-case values. The adapter currently assumes that the result columns are always named DATABASE, SCHEMA, NAME, and TYPE.

Because Agate column selection is case-sensitive, this causes list_relations_without_caching to fail when the driver returns names such as database, schema, name, and type.

Changes

• Build a case-insensitive mapping of the result column names.
• Select the original column names returned by the driver.
• Raise a descriptive CompilationError when one or more expected metadata columns are missing.
• Add a unit test covering uppercase, lowercase, and mixed-case column names.

Reproduction

The issue occurs when the result of the list_relations_without_caching macro has column names such as:

database
schema
name
type

instead of:

DATABASE
SCHEMA
NAME
TYPE

The existing implementation calls:

results.select(["DATABASE", "SCHEMA", "NAME", "TYPE"])

which fails because Agate column selection is case-sensitive.

Testing

tox -e py311 -- tests/unit/test_adapter.py -k list_relations