Use the shared aggr_test_schema in CSV scan tests instead of a local copy #26306

Is your feature request related to a problem or challenge?

The test module in core/src/datasource/physical_plan/csv.rs defines a private aggr_test_schema() that is a verbatim copy of the public datafusion::test_util::aggr_test_schema(). If the shared schema changes (for example new columns or metadata on aggregate_test_100.csv), the CSV tests keep using the stale copy.

The same module also repeats about 35 lines of scan setup in every rstest case (session context, partitioned_file_groups, CsvOptions, TableSchema, CsvSource, FileScanConfigBuilder), differing only in projection, limit and the expected snapshot (L130-L442). json.rs (L213-L372) does the same with prepare_store and JsonSource.

Describe the solution you'd like

1. Delete the local aggr_test_schema in the csv.rs tests and import crate::test_util::aggr_test_schema. Keep aggr_test_schema_with_missing_col as a derivation of it.
2. Optionally add a test helper such as csv_scan_config(file_compression_type, file_schema, |builder| ...) -> FileScanConfig that does the shared setup, so each rstest case only applies its projection or limit and asserts. Do the same for json.rs.

Describe alternatives you've considered

Do step 1 only.

Additional context

Step 1 is a small, self-contained change and a good first issue. Line references are pinned to commit 0d79252.