Deepeval Hallucination Prompt

Reason prompt for the DeepEval hallucination metric.

gllm-evals reports hallucination as the share of contexts the output contradicts (lower is better), the direction deepeval used before 4.2.0. This is deepeval 3.9.9's reason prompt, so the judge explains the score gllm-evals returns instead of deepeval's flipped one.