add streaming OLS implementation with duckdb arrow interop

[apoorvalal]

No description provided.

[Copilot]

Pull Request Overview

This PR introduces a streaming OLS implementation that leverages DuckDB's Arrow IPC support for memory-efficient regression on large datasets. The implementation uses sufficient statistics to process data in chunks rather than loading everything into memory.

• Adds streaming regression functionality with DuckDB Arrow interop
• Implements sufficient statistics-based OLS and Ridge regression
• Provides comprehensive test coverage for the streaming implementation

Reviewed Changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 3 comments.

File	Description
duckreg/stream.py	Core streaming regression implementation with Arrow IPC support
tests/test_stream.py	Test suite covering streaming regression functionality
duckreg/init.py	Exposes StreamingRegression in public API

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Using mutable default arguments (None) should be replaced with proper Optional type annotations and default to None explicitly. The type hints should be Optional[list[str]] and Optional[str].

Suggested change

	feature_cols: list[str] = None,
	target_col: str = None,
	feature_cols: Optional[list[str]] = None,
	target_col: Optional[str] = None,