Remove support for benchmarking with the Tier 2 interpreter

[mdboom]

We haven't been optimizing the tier 2 interpreter (without the JIT) for performance in a long time. We haven't included it in the weekly benchmarks for at least a year.

I think we should remove the option to benchmark with the tier 2 interpreter, because it would free up the limited space of 10 options for other things. I just added a box for tail calling, putting us at the maximum number of options.

@markshannon: Is this ok for you, as the biggest user of the tier 2 interpreter?