sv-benchmarks unsoundness pre-SV-COMP 2025

[sim642]

• [x] nla-digbench-scaling/geo1-ll2_unwindbound1 with no-overflow property: something to do with unrolling, because disabling loopUnrollHeuristic makes us sound again. Incorrect expected verdicts? https://gitlab.com/sosy-lab/benchmarking/sv-benchmarks/-/issues/1415, https://gitlab.com/sosy-lab/benchmarking/sv-benchmarks/-/merge_requests/1558.
• [x] termination-restricted-15/IntPath with termination property: the verdict is technically right, but we have "Both branches dead" (#1576). This also has something to do with unrolling, because disabling loopUnrollHeuristic makes us sound again. Issue from Apron normalization: #1585.