This partially addresses #259 by supporting reconstructing of probability distributions in the special case where the circuit does not separate into disconnected components.

I plan to maintain this as a fork until we figure out how it should integrate with the existing code. (There might be some refactoring that would be appropriate before we integrate it.)

Demonstration notebook here.

I expect coverage to fail because I have not added tests (there's only the notebook for now), but adding tests should be pretty straightforward.

Remaining action items

• _decompose_qpd_measurements should throw an error if circuits already has a register called "qpd_measurements"
• update reconstruction for samplerv2
• change to 4-qubit GHZ state in the tutorial, perhaps
• add the tutorial to README.rst
• fix cut_wires to be compatible with classical registers