Adding benchmark circuits with arbitrary gatesets

[jordandsullivan]

From the very inception of UCC (see Discussion from Sep 2024), we've considered the tradeoffs of decomposing circuits in order compare different compilers against each other.

Up until this point, we have made the decision to start our compiler benchmarking by first decomposing input circuits into a standard gateset, which makes it easy to compare the number of multi-qubit gates before and after compilation.

However, as we begin to test out the performance of existing open-source tools designed for gate synthesis rather than pure optimization of quantum circuit DAGs, like Synthetiq and potentially GUOQ/WISQ, we may wish to start comparing performance on circuits with arbitrary gates, which may be more realistic in terms of quantum programs in the wild.

For these types of benchmarks, because approximate synthesis is often utilized, we will also want to determine what an acceptable threshold is for circuit equivalence.

[willzeng]

I'd support this. Matches the main programmer use case where usually the algorithm is defined in some arbitrary gate set first