No noise model reset

[antalszava]

Context
Upon device initialization, a noise model can be passed to a Qiskit Aer device by specifying the noise_model keyword argument. The noise model is then stored as an attribute, which is initialized to None and is further reset to None once dev.reset is called.

PennyLane core calls dev.reset in certain places (e.g., as part of QNode evaluation) which eventually result in "losing" the user defined noise model.

Furthermore, Qiskit Aer devices persist the options that were set for them prior. This affects consecutive device creations: e.g., devices created after a noisy device creation would also have noise be set as an option.

Changes

• This PR removes the noide_model attribute and moves the dev.backend.set_options call directly to QiskitDevice.__init__ from QiskitDevice.run.
• Adds a call to the clear_options method for Aer backends

[josh146]

Looks good @antalszava, however it seems that this is causing a large number of tests to fail 🤔

Are some of the tests implicitly written assuming noise models are being reset?

[josh146]

nice test :)

[glassnotes]

I get two failing tests (running 0.13.0 dev for both PennyLane and PennyLane-Qiskit):

=========================================================================================================================== short test summary info ===========================================================================================================================
FAILED test_ibmq.py::test_account_error - Failed: DID NOT RAISE <class 'qiskit.providers.ibmq.exceptions.IBMQAccountError'>
FAILED test_sample.py::TestTensorSample::test_hermitian[AerDevice-qasm_simulator-8192-False-1.0-1.0-1.0] - AssertionError: assert False

But neither of these seem to involve the noise model (at least not directly).

In terms of the actual fix - my code that was previously not working due to the noise model being reset is now working as expected; the noise model persists between jobs.

[codecov]

Codecov Report

Merging #114 (4d7899a) into master (7ba24ac) will increase coverage by 0.34%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #114      +/-   ##
==========================================
+ Coverage   97.92%   98.26%   +0.34%     
==========================================
  Files           7        7              
  Lines         289      289              
==========================================
+ Hits          283      284       +1     
+ Misses          6        5       -1     

Impacted Files	Coverage Δ
pennylane_qiskit/qiskit_device.py	98.13% <100.00%> (+0.62%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7ba24ac...4d7899a. Read the comment docs.

[josh146]

@antalszava not sure if I missed it, but don't forget to update the changelog!

[antalszava]

@josh146 thanks for the reminder :) Added it in #116.