Add buggy benchmarks

[konnov]

Currently, all the tests are assumed to report NoError. Add configurations, where errors are expected.

[shonfeder]

Any preference on the priority here? If we get apalache-mc/apalache#86 in place first (perhaps also providing an option to output JSON, if that would be useful), then we can overhaul the current text munging to just use simple data processing on the resulting serialized output. That should help clean up the scripts a bit, and be more robust in the long term.

[konnov]

It is not really a priority. But let's not forget about it, as we do not want unexpected false positives.