[python] Clean up normal files when composite writer metadata creation fails

[wangzhigang1999]

Purpose

Fixes #9916

DataVectorWriter and DedicatedFormatWriter can leave a normal data file behind if writing succeeds but a later step, such as reading the file size, fails before returning its metadata.

Wrap normal-file output and metadata creation in try/except. On failure, call delete_quietly for the current file and re-raise the original exception.

Tests

Add one regression test covering both writers: write a real Parquet file, fail while reading its size, and assert that no Parquet file remains. Both cases fail on the base revision and pass with the fix.

The new regression and existing composite flush/shared-shredding tests passed (10 test methods). Flake8 and whitespace checks passed with Python 3.12.6 / PyArrow 19.0.1.

[Copilot]

🟢 Approval recommended

The implementation directly addresses the reported leak and includes focused regression coverage.

Pull request overview

Adds cleanup for normal data files when composite writer metadata creation fails.

Changes:

• Wraps normal-file writing and metadata creation in cleanup handlers.
• Adds regression coverage for both composite writers.

File summaries

File	Description
dedicated_format_writer.py	Deletes failed normal-file output and re-raises errors.
data_vector_writer.py	Adds equivalent cleanup handling.
writer_file_cleanup_test.py	Verifies failed metadata creation leaves no Parquet file.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[JingsongLi]

Reviewed the normal-file cleanup path. Deleting the completed data file when subsequent metadata creation fails prevents orphan files while preserving the original exception; the regression coverage exercises both composite writers. No additional blocking issues found.