Support direct Arrow RecordBatch to Parquet conversion

## Search before asking
- [x] I searched in the [issues](https://github.com/apache/fluss/issues) and found nothing similar.

## Description
This issue tracks the Arrow-to-Parquet conversion part of splitting parent task #437.

We need a reusable utility or writer path that can convert Arrow `RecordBatch` directly into Parquet without first converting through row-oriented representations. This would improve the efficiency and clarity of Arrow-native data flows, especially for tiering and other lakehouse-related write paths.

Possible scope:
- provide a reusable Arrow `RecordBatch` to Parquet conversion utility;
- define the supported Arrow/Parquet type mapping and failure behavior;
- make the conversion path suitable for direct reuse by tiering writers and other lake components.

This is intended to be one sub-task of #437, while enabling tiering source to read data as Arrow `RecordBatch` is tracked separately.

## Willingness to contribute
- [ ] I'm willing to submit a PR!

Start by reading parent issue #437 and tracing the existing tiering and lakehouse write paths to find where Arrow RecordBatch and Parquet handling belong. Define the supported Arrow-to-Parquet type mapping and failure behavior, then confirm the reusable direct conversion path works for tiering writers and other lake components.

Written by the indexing model from the issue text.