BAD: BiAs Detection for Large Language Models in the context of candidate screening