[SPARK-56584][PYTHON] Generalize RESULT_TYPE_MISMATCH_FOR_ARROW_UDF error class and remove dead SCHEMA_MISMATCH_FOR_ARROW_PYTHON_UDF

[Yicong-Huang]

What changes were proposed in this pull request?

Two related cleanups in the PySpark result-verify path:

1. Rename error class RESULT_TYPE_MISMATCH_FOR_ARROW_UDF to the more general RESULT_COLUMN_TYPES_MISMATCH (parallel to RESULT_COLUMN_NAMES_MISMATCH / RESULT_COLUMN_SCHEMA_MISMATCH). The error is raised from the generic verify_arrow_result path in python/pyspark/worker.py; the name shouldn't mention "ARROW_UDF".
2. Reword the message to align with its siblings:
   • Before: Columns do not match in their data type: <mismatch>.
   • After: Column types of the returned data do not match specified schema. Mismatch: <mismatch>.
3. Remove the dead error class SCHEMA_MISMATCH_FOR_ARROW_PYTHON_UDF. git grep confirms no code path raises it, and its message body is identical to SCHEMA_MISMATCH_FOR_PANDAS_UDF.

Why are the changes needed?

Part of SPARK-55388 (Refactor PythonEvalType processing logic). Cleanup to make error class names and messages consistent across the result-verify path, and to remove dead code.

Does this PR introduce any user-facing change?

Yes. User-visible error class name and message for result column type mismatches in Arrow UDFs change. The unreleased SCHEMA_MISMATCH_FOR_ARROW_PYTHON_UDF class is removed (no code raised it).

How was this patch tested?

Updated 4 existing asserts in test_arrow_grouped_map.py and test_arrow_cogrouped_map.py that match the new message.

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

No

[Yicong-Huang]

cc @gaogaotiantian

[HyukjinKwon]

Merged to master.