Refactor OpenMLEvaluation to use dataclass (Fixes #1540)

[Sumitb09]

Fixes #1540

Summary

Refactored the OpenMLEvaluation class to use Python's @dataclass decorator to simplify the class implementation and reduce boilerplate code.

Changes

• Converted OpenMLEvaluation into a @dataclass.
• Replaced manual attribute assignments with dataclass field definitions.
• Used asdict() for dictionary conversion in _to_dict.
• Added default values for optional fields (value, values, and array_data) to maintain compatibility with existing usage.
• Ensured type hints follow modern Python standards (list instead of typing.List).

Benefits

• Cleaner and more maintainable code.
• Reduced boilerplate for attribute initialization.
• Improved readability and type safety.

Testing

• Ran black for formatting.
• Ran ruff check --fix for linting.
• Executed full test suite with pytest.

All tests pass successfully.

Before

The class relied on manual initialization and attribute assignments.

After

The class now uses Python's dataclass functionality, automatically generating __init__, __repr__, and other utility methods.