Decomposition of gates with 3 or more qubits

[garrison]

We may eventually want to perform gate cutting on gates that act on 3+ qubits: for instance, CSwapGate, CCXGate, C3XGate, etc. In each case, we could start by only support cutting such a gate into two pieces, as that is complex enough to handle and will cover the main efficiency gains that could come from doing this.

• The three-qubit gates could be split one of two ways: either ABB or AAB.
• The CNXGates can be split in N - 1 different ways, depending on how many control bits end up split on the same side as the target bit.

Of course, in addition to refactoring QPDBasis, we'll need to derive what the coefficients of the decomposition are in each of the above cases we want to support.

[garrison]

There are decompositions of the Toffoli gate in Remark 7.1 of https://arxiv.org/abs/2312.11638v2. The split can either be AAB (two control bits are local to each other) or ABB (the target bit is local to only one control bit). In each case, one additional qubit is needed on the side of the cut with two qubits, and two Toffoli gates among the three resulting bits (see the paper for details).