Remove subsystem_labels option from DynamicsBackend

[DanPuzzuoli]

Summary

Closes #235

Details and comments (out of date, see next comment)

This is currently a draft to prototype a potential solution to #235, in which the subsystem_labels option is removed, and all qubits removed via the subsystem_list argument of from_backend are actually "kept" as trivial 1-dimensional systems.

So far:

• I've commented out all lines/blocks using subsystem_labels.
• Where necessary, new code has been added, e.g. construction of subsystem_dims in from_backend has been changed to include all pre-existing subsystems, but setting the dimension of the "removed" subsystems to 1 (instead of not including them at all).

Initial testing in a jupyter notebook seems to indicate that this behaves as expected:

• I can do from_backend with a non-trivial subsystem_list, and the backend is built without issue.
• I haven't yet done any verifications of correctness, but it seems like we can run sims and compute measurement results as well.

To do:

• Add number of qubits to the backend wherever it makes sense.
• Do some initial verification that the calculations are correct.
• The subsystem_labels option was used to raise errors of non-existent subsystems were measured. As this no longer exists, it would be nice to still warn users if their schedule measures a trivial subsystem.
  • This could just be done by checking the dimension of the measured subsystem.
• Pre-existing tests need to be modified.

Question:

• This will remove the subsystem_labels option. I'm not sure how this could be deprecated, because it will literally serve no purpose and have no effect anymore.

[DanPuzzuoli]

I've now implemented all of the above changes. Here is the summary:

• subsystem_labels has been removed from the DynamicsBackend options. Now, we will always view the labels of subsystems in DynamicsBackend as being of the form [0, ..., num_qubits - 1].
• This option was used in DynamicsBakend.from_backend - the argument subsystem_list was stored in subsystem_labels to keep track of which subsystems were kept. With this being removed now, the returned DynamicsBackend instead "keeps" the removed subsystems (the returned backend still looks like it has all qubits), but treats them as trivial by listing their dimension as 1 in the subsystem_dims argument.

Luckily this change doesn't impact any of the numerical functionality. In addition to the above:

• Error checking/raising involving subsystem_labels have been changed. E.g. now, rather than checking if an index is in subsystem_labels it will check if it's < len(subsystem_dims).
• A warning has been added if a trivial subsystem (with dimension 1) is being measured.
• I've modified and added tests to check for the above behaviour.

I've also added an upgrade release note explaining this change.