mlchrzan/pairadigm: [0.5.4] 10/10 on the Splits

Added

• Users now have the ability to pass their own system_prompts and comparison_prompts if desired.

Updated

• score_items() now also can respect the splits created when generating pairings (or providing split data).

Fixed

• Syntax error where a print statement was misplaced in score_items() causing the method to not function.
• Typo fixed: tempature → temperature in the signature, docstring, call site inside generate_breakdowns_from_paired(), and the build_pairadigm() call.
• Index mismatch: Changed results from a list to a dict, and replaced r[0] for r in results/[r[1] for r in results] with result_df.index.map(...) so non-zero-based or non-contiguous DataFrame indices are handled safely.
• Duplicate llm_annotator_cols entries: Added an if decision_col not in self.llm_annotator_cols guard before appending.