feat: flatten_stream returns list[Path] (v2.0.4)

[scottdraper8]

Summary

• flatten_stream now returns list[Path] (the paths of all files written) instead of None
• Return value propagates up through writer.close() → stream_process() → flatten_stream()
• Fully backwards-compatible — callers that discard the return value are unaffected

Changes

• Writers (base.py, csv.py, arrow_base.py, avro.py): StreamingWriter.close() abstract signature updated to -> list[Path]; each concrete writer tracks and returns written file paths
• Streaming (streaming.py): stream_process() captures and returns list[Path] from writer.close()
• API (api.py): flatten_stream() captures and returns result of stream_process(); docstring updated with Returns: section
• Tests: Updated assert result is None assertions to verify returned list is non-empty and all paths exist
• Docs: Updated api.md and streaming.md with new return type and usage examples

Test plan

• poetry run pytest tests/unit/test_api_flatten.py::TestFlattenStream -v
• poetry run pytest tests/unit/test_streaming.py -v
• poetry run pytest (full suite)

---

Note

Low Risk
Low risk: behavior is additive for callers who ignored the return value, and changes are limited to return-type plumbing and path tracking in streaming writers.

Overview
flatten_stream() now returns a list[Path] of all output files written (rather than None), by propagating the return value from StreamingWriter.close() through stream_process() to the public API.

To support this, streaming writers (CSV, PyArrow/Parquet/ORC, Avro) now track written file paths and return them on close(), and docs/tests were updated accordingly; version is bumped to 2.0.4.

^{Written by Cursor Bugbot for commit a5f8440. This will update automatically on new commits. Configure here.}