Deprecate old QNode

[antalszava]

Context:
Currently, the refactored QNode is in the pennylane/beta folder and users could potentially use it through referencing there. However, maintaining two types of QNode objects is not favorable in the long run. Therefore, deprecation of the old-style QNode is needed.

Description of the Change:
Deprecates the old QNode such that only the new version is supported. Moves all related files from the beta folder.

Benefits:
Only the new QNode version needs to be maintained.

Possible Drawbacks:
New syntax to be used e.g. for the decorator, several changes need to be noted.

Related GitHub Issues:
Series of previous QNode PRs, such as #420, #425

[codecov]

Codecov Report

Merging #440 into master will decrease coverage by 0.09%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master     #440     +/-   ##
=========================================
- Coverage   99.33%   99.24%   -0.1%     
=========================================
  Files          46       45      -1     
  Lines        3760     3303    -457     
=========================================
- Hits         3735     3278    -457     
  Misses         25       25

Impacted Files	Coverage Δ
pennylane/qnodes/device_jacobian.py	90.9% <ø> (ø)
pennylane/beta/__init__.py	100% <ø> (ø)	⬆️
pennylane/interfaces/autograd.py	100% <ø> (ø)
pennylane/qnodes/__init__.py	100% <ø> (ø)
pennylane/qnodes/cv.py	100% <ø> (ø)
pennylane/beta/plugins/expt_tensornet_tf.py	92.1% <ø> (ø)	⬆️
pennylane/qnodes/decorator.py	100% <ø> (ø)
pennylane/qnodes/qubit.py	100% <ø> (ø)
pennylane/interfaces/__init__.py	100% <ø> (ø)
pennylane/qnodes/jacobian.py	100% <100%> (ø)
... and 9 more

---

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 375d25a...f57ec37. Read the comment docs.

[josh146]

This is great Antal, thanks for helping finish this! I can't seem to compile the documentation locally, so I'm going to see if I can fix that quickly. Once done, this should be ready for merging.

Before I forget, let's remember to update the changelog as well 😆

[josh146]

Update: I have the docs correctly compiling, and viewable here

[antalszava]

That is really nice @josh146 , thanks!

A really nasty caveat that I had to double-check (my latest commits include a correction to these cases) is plainly exchanging a circuit with a OperationRecorder the following way:

def circuit(weights):
    CVNeuralNetLayers(*weights, wires=range(4))
    return qml.expval(qml.X(wires=0))

with qml.utils.OperationRecorder() as rec:
    CVNeuralNetLayers(*weights, wires=range(4))
    return qml.expval(qml.X(wires=0))

Extra care needs to be taken, as in the second case the test will simply pass (due to it reaching the return statement) without really checking the test case itself (as most of the times this code snippet would be placed at the beginning of the test).

This is the reason why the return statements were (and should be) removed when using the OperationRecorder.

Just thought it's worth pointing this out just as a future use of the OperationRecorder.