Implement optimal cutting of a single, general two-qubit unitary

[garrison]

The decomposition we currently use for cutting a general two-qubit unitary is given by https://arxiv.org/abs/2006.11174. When that work was published, it was not known whether their decomposition was optimal. More recently, the authors of https://arxiv.org/abs/2312.11638 provided a better decomposition that is provably optimal (see e.g. the intro to Sec. 3 in the current version of their paper). However, the decompositions are equivalent for all the "named" gates we care about, such as CRXGate, and even XXPlusYYGate. We should eventually implement this optimal decomposition, but it will only be relevant for arbitrary UnitaryGates and similar. All the standard gates that I have considered are already cut with optimal overhead by the existing method.

[garrison]

The optimal decomposition will require an ancilla on each side of the cut. See https://github.com/Qiskit-Extensions/circuit-knitting-toolbox/pull/568#issuecomment-2088453019.