[WIP] Generate json files for tests (which used pickled data)

[dekool]

Summary

changed the pickle files used in the fitter tests into json files.
in addition, reordered the comparing of the real results and the expected results into one function.

Details and comments

• in order to create the json files I loaded the pickle files and saved the data inside them as json files (using the "save_results_as_json" function I added in the utils).
• in addition, I didn't changed the tests results, just splited between the tests settings and expected results for easier change of the way the expected results are kept (I will save them also in a json format)
• I will add a code that generate the expected results and saves them as a json

I also checked the sizes of the json files compared to the pickle files, and found that -

• for small files (less than 100KB) - the json files are larger in around 2KB
• for large files - the json files are a bit smaller than the pickle files

the run times of the tests are almost the same with pickle files and json files (when running them on my laptop) - with the json files the average was a little faster (in the scale of only 1/100 sec)

[CLAassistant]

All committers have signed the CLA.

[chriseclectic]

Hi @dekool thanks for starting this. Can you have a look at the discussion and final merged version of PR #387 for how to add a script for re-generating the reference JSON files.

[ShellyGarion]

It will be better to merge this only after #407 is merged.
So that the json files with the results will be generated using the new Clifford / CNOTDihedral classes.
In addition, there will be an option to add a random seed in randomized_benchmarking_seq.

[chriseclectic]

@ShellyGarion if you do that can we integrate this PR with those so we can test them. Currently several tests are disabled for RB (#402) so we can't evaluate that those PRs are working correctly.

[ShellyGarion]

should it be coherent_noise_model?

[ShellyGarion]

cnot_dihedral does not belong here...

[ShellyGarion]

maybe you mean 'original_ydata' ?

[ShellyGarion]

Note: the generation of RB json files depends on #407

Hence the code in this PR should be revised after #407 is merged.

1. The Cliffords functions for random / decompose were updated
2. There should be a fixed seed for generating the RB sequences
3. The API for interleaved RB was updated

[yaelbh]

1. Is the code in the generate_data.py files a duplicate of code that lies somewhere else? If yes then the duplication should be removed. If no then how did you decide how to write it?
2. I recommend to split to three PRs, for measurement calibration, quantum volume, and RB. It's ok to have the same change in utils.py in all three PRs. This way only the RB pull request will depend on another non-merged RB pull request.

[dekool]

1. most of the code is inspired by different qiskit notebooks explaining the relevant subjects. I can't call functions of the notebooks from the code, so I can't remove the duplicated parts easily. there are also some places which I used non-noisy tests of the same functionality as basis. at these places I can remove the duplication from the common part, but it will cause dependency between different tests
2. good idea. I will split this PR into 3

[ShellyGarion]

Generally, you can also follow this example:
https://github.com/Qiskit/qiskit-ignis/blob/master/test/accreditation/generate_data.py

[dekool]

this PR was split into 3 PR in order to separate data generation of different functionalities -
RB - #444
QV - #445
measurement calibration - #446