Update InterpBench ioi models

[cybershiptrooper]

IOI and IOI next token are actually the same model:

# load both cases
from circuits_benchmark.utils.ll_model_loader.ll_model_loader_factory import get_ll_model_loader
from circuits_benchmark.utils.get_cases import get_cases
import torch

cases = get_cases(indices=["ioi", "ioi_next_token"])
ll_model_loader = get_ll_model_loader(cases[0], interp_bench=True)
corr, model = ll_model_loader.load_ll_model_and_correspondence(device="cpu")
ll_model_loader = get_ll_model_loader(cases[1], interp_bench=True)
corr2, model2 = ll_model_loader.load_ll_model_and_correspondence(device="cpu")
# check if both models have same weights

for p1, p2 in zip(model.parameters(), model2.parameters()):
    assert torch.all(p1 == p2)  # this should not happen