[WIP] Restless experiments data processing

[catornow]

ToDo:

• Add tests.
• Add additional tests.
• Improve tutorial.
• Improve summary and details.

Summary

This PR implements the new data processing nodes RestlessNode and RestlessToCounts. These allow users to perform and analyze experiments based on restless measurements (without qubit reset) to significantly speed-up calibration experiments compared to standard experiments based on active reset, see https://arxiv.org/pdf/2202.06981.pdf.

Details and comments

Certain experiments, e.g., calibration experiments or Randomized Benchmarking, can be run in a restless setting.
In a restless experiment the qubits are not reset after a measurement and the subsequent circuit is run after a short delay of, e.g., 1 us. Consequently, the initial state of a circuit is either |0> or |1> depending on the outcome of the previous circuit. This makes the post-processing of the single measurement outcomes of each circuit necessary which is done by the new data processing nodes introduced in this PR. Crucially, restless experiments provide a fast alternative to several characterization and calibration tasks based on active reset.

The following run options define a restless experiment:

1. Set the rep_delay to a small value, e.g. 1 us.
2. Ensure that the qubits are not reset to the ground state throughout the experiment: init_qubits=False.
3. Set memory=True to ensure that the single measurement shots are returned. This is needed for post-processing.
4. Set meas_level=2.

Restless experiments will automatically be post-processed by the new restless nodes.

Example

We can define a restless RoughDrag experiment by specifying the restless run options:

restless_cal_drag = RoughDrag(qubit, ...)
restless_cal_drag.set_experiment_options(...)

# set restless run options.
restless_cal_drag.set_run_options(rep_delay=1e-6, meas_level=2, memory=True, init_qubits=False)

[eggerdj]

This is looking good. Can you setup an extra test that uses a restless simulator to produce counts for an experiment so that we can create an end-to-end test?

[eggerdj]

Looking good. We must also document the arguments that get_processor needs to extract from the metadata in its docstring.