BUG: get_all_harm_metrics finds nothing for SEXUAL_CONTENT and REPRESENTATIONAL #3542

The sexual and representational Likert scorers (and the Azure sexual category) use the canonical names SEXUAL_CONTENT and REPRESENTATIONAL as their harm_category. find_harm_metrics_by_eval_hash maps those to sexual_metrics.jsonl and representational_metrics.jsonl, but get_all_harm_metrics just builds "{harm_category}_metrics.jsonl", so it looks for files that don't exist and returns an empty list.

from pyrit.score import get_all_harm_metrics
len(get_all_harm_metrics("SEXUAL_CONTENT"))    # 0
len(get_all_harm_metrics("sexual"))            # 5
len(get_all_harm_metrics("REPRESENTATIONAL"))  # 0