Extend `generate_xyz_hamiltonian` to allow for arbitrary coefficients.

[BryceFuller]

What should we add?

Right now generate_xyz_hamiltonian allows one to specify the coupling coefficients across all edges for each interaction type in the Hamiltonian, but for some applications one might want to, for example, use a set of randomly sampled coefficients for the XX terms.

This could be supported by accepting a dictionary mapping edges in the coupling map to coefficients.

For example:

np.random.seed(42)
xx_coeff_dict = {edge: np.random.randn() for edge in coupling_map}

# Get a qubit operator describing the Heisenberg XYZ model
hamiltonian = generate_xyz_hamiltonian(
    coupling=coupling_map,
    coupling_constants=(xx_coeff_dict, np.pi / 4, np.pi / 2),
    ext_magnetic_field=(np.pi / 3, np.pi / 6, np.pi / 9),
)

A few sanity checks on the input are that if a dict is passed in for one of the coeffs:

• All edges in the coupling graph must be present in the dictionary
• Each edge should appear only once, i.e if (0,1) and (1,0) appear and have different coeffs, throw an error