basis translation method based on unitary synthesis

[ajavadia]

As a follow-up to #4446, we should have another method for basis translation that relies on OneQubitEulerDecomposer and TwoQubitBasisDecomposer.

There are some advantages to this, the main one being that the decompositions are algorithmic and can cover a wider range of basis. For example if someone comes up with some esoteric two-qubit gate as the native gate, this code will handle it whereas the EquivalenceLibrary is only as good as its populated content.

This has to be done after all 3+ qubit gates are decomposed. Then I think a sequence of 1-qubit and 2-qubit block collection passes followed by invoking those synthesizers will do the job.

It remains to be seen which basis translation method works best and it's something we will need to benchmark. But this empowers us to do so.