✨ Support for synthetic benchmarks

What's the problem this feature will solve?

In some discussions over the past couple of weeks, the interest in synthetic quantum circuit benchmarks has come up.
An example for such benchmarks would be https://github.com/qu-tan-um/QUEKO-benchmark (based on https://arxiv.org/abs/2002.09783), which recently got extended in https://arxiv.org/abs/2502.08839
These benchmarks are particularly tailored to specific architectures.
They are specified in terms of native gates of the device and come with a known bound for the optimal depth in which they can be realized on the architecture.

Describe the solution you'd like

Ideally, we would adapt the generation code so that we can generate synthetic circuits for arbitrary architectures.
Since these benchmarks do not really make sense at the algorithm or the target-independent level, some special thought probably needs to be put into how to best integrate such benchmarks.
It would also be great, if we could somehow expose the known optimal solution.

No files, tests, or entry points are named in the issue. Start by locating the existing circuit-generation and benchmark-integration code, then determine how architecture-specific circuits and known optimal solutions should be represented. Done would require an agreed design and support for arbitrary architectures.