[VL] Respect caseSensitiveAnalysis in IcebergScanTransformer metadata-column handling

[sabbasani]

This PR fixes case-sensitive column-name handling in IcebergScanTransformer.

IcebergScanTransformer used unconditional .toLowerCase(Locale.ROOT) when handling schema and input-file-related metadata column names. This does not respect Spark's spark.sql.caseSensitive configuration and can cause mixed-case user columns to collide with metadata-related column names when spark.sql.caseSensitive=true.

This PR replaces the unconditional lowercasing with the existing canonical helper:

ConverterUtils.normalizeColName

normalizeColName respects SQLConf.get.caseSensitiveAnalysis:

when spark.sql.caseSensitive=true, the original column casing is preserved;
when spark.sql.caseSensitive=false, column names continue to be lowercased as before.

What changes are proposed in this pull request?

How was this patch tested?

Added regression coverage for both the general case-sensitive execution path and the Iceberg-specific path.

FallbackSuite

Tests cover:

case-sensitive top-level column access with native execution;
native aggregation with GROUP BY;
native join execution;
default case-insensitive behavior remaining unchanged.

Results:

FallbackSuite
21 tests
21 passed
0 failed
0 errors

The tests compare Gluten results against vanilla Spark and also verify that native Gluten/Velox transformer nodes are present.

VeloxIcebergSuite

Added coverage for:

Iceberg scan with spark.sql.caseSensitive=true;
input_file_name() with case-sensitive mode enabled;
a mixed-case data column such as Input_File_Name not being confused with metadata-related columns;
case-sensitive metadata-column handling.

Results:

VeloxIcebergSuite
19 tests
19 passed
0 failed
0 errors

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

Yes, AI assistance.

[github-actions]

Run Gluten Clickhouse CI on x86

[github-actions]

Run Gluten Clickhouse CI on x86