refactor citation score filtering

[sumukshashidhar]

Summary

• refactor citation score filtering pipeline
• add dataclasses and type hints
• improve logging and error handling

[alozowski]

Can we try to simplify this with StageConfig(**cfg) for brevity? This can raise a TypeError if any expected key is missing, but it can be a better way to delegate default values to the dataclass and keep the logic cleaner, if we're confident the config is well-formed

Kinda like this:

def _get_stage_config(config: dict[str, Any]) -> StageConfig:
    return StageConfig(**config.get("pipeline", {}).get("citation_score_filtering", {}))