Implement native support for ORC InsertIntoHiveTable writes

[weimingdiit]

Describe
Auron's native write path currently supports Parquet InsertIntoHiveTable, but ORC-backed InsertIntoHiveTable operations still fall back.

This prevents native acceleration for Hive ORC insert workloads even when the input side is already native.

Describe the solution you'd like
Add native support for ORC InsertIntoHiveTable writes.

The implementation should preserve Spark semantics for:

• insert overwrite vs append
• partition handling
• dynamic partition ordering
• output correctness and write statistics

Regression coverage should include:

• non-partitioned ORC insert
• partitioned ORC insert
• overwrite and append cases