Refactor reconstruct_expectation_values

[caleb-johnson]

As a follow-up to #385 and in the larger effort associated with #363 , this PR refactors reconstruct_expectation_values to accept the outputs from the Sampler primitives directly (with the assumption that the metadata contains QPD bit information).

This PR will also refactor execute_experiments to output a data structure which can also be consumed by reconstruct_expectation_values. While this will be a breaking change, it will allow us to prevent an abrupt removal of a core function. Additionally, this breaking change actually results in returning more info to the user from execute_experiments. The user will now have the entire SamplerResult object(s), including the QPD bit information; whereas, before they only got the output quasi-dists and qpd bit info.

[garrison]

Thanks, this is looking pretty great already. Right now, it's a net loss of 13 lines 🎉. I will comb through it more carefully shortly.

[caleb-johnson]

Thanks, this is looking pretty great already. Right now, it's a net loss of 13 lines 🎉. I will comb through it more carefully shortly.

Yes, and most of the added lines are documentation and tests; whereas, most of the removed lines are source code. Hooray

[garrison]

Thanks -- this looks great. Final tally: net addition of 23 lines, even when counting docstrings, new tests, and the release notes 🎉.