fix(evaluator): guard IndexError on truncated judge choice (JudgeEvaluator/RMBEvaluator)

[WatchTree-19]

problem

JudgeEvaluator.score and RMBEvaluator.score extract the winning model letter with:

choice = prediction.split("\"Choice\": \"Model ")[-1][0] if len(prediction) != 0 else None

the guard checks that prediction is non-empty, but the [0] indexes the split tail, which can be empty even when the prediction isn't - e.g. a judge response truncated right at the marker (... "Choice": "Model with nothing after, common under max_tokens clipping). then split(...)[-1] is "" and ""[0] raises IndexError, crashing scoring for the whole batch.

repro

p = '{"Choice": "Model '            # truncated before the letter
p.split('"Choice": "Model ')[-1][0]  # IndexError: string index out of range

fix

index the split tail only when it's non-empty (missing choice -> None, which simply doesn't match the gold winner). applied to both JudgeEvaluator and RMBEvaluator.

test

added tests/openicl/test_icl_judge_evaluator.py: the truncated-marker case no longer crashes (scores 0.0), and a normal "Choice": "Model A" still scores correctly.