[WIP] Make a separate base quantum tape class

[mariaschuld]

For future development, for example when introducing a batch tape, it would be cleaner to have a "base" quantum tape class that only implements functionality for constructing, evaluating and validating circuits. The current QuantumTape class would inherit from this base class and add functionality for differentiation.

The aim of this PR is to check and discuss how invasive such a change would be.

About naming: we reserve the name QuantumTape for the new base tape class, and introduce JacobianTape for the tape that implements differentiation. JacobianTape has all functionality of the current QuantumTape class. While this is much clearer, it makes the PR quite large, even though most changes are just renamings.

Advantages: Conceptually, the representation of a circuit and the ability to compute its Jacobian are two separate features, one extending the other. Classes can now inherit from the base tape without inheriting a specific implementation of the jacobian method.

Disadvantages: Differentiation and circuit construction are not entirely distinct, since the circuit construction uses logic on differentiable and non-differentiable parameters. It seems that the only function that lives at this intersection is _update_gradient_info(), which adds the preferred diff method to trainable parameters. Here, this function is made a property of the jacobian quantum tape.

Note: At the moment, the PR does not contain any changes to documentation, which would have to be added if we decide to follow this route.

[codecov]

Codecov Report

Merging #829 into master will increase coverage by 0.00%.
The diff coverage is 99.10%.

@@           Coverage Diff           @@
##           master     #829   +/-   ##
=======================================
  Coverage   97.68%   97.68%           
=======================================
  Files         135      136    +1     
  Lines        9107     9117   +10     
=======================================
+ Hits         8896     8906   +10     
  Misses        211      211           

Impacted Files	Coverage Δ
pennylane/tape/interfaces/autograd.py	100.00% <ø> (ø)
pennylane/tape/interfaces/tf.py	96.22% <ø> (ø)
pennylane/tape/interfaces/torch.py	100.00% <ø> (ø)
pennylane/tape/tapes/cv_param_shift.py	99.20% <ø> (ø)
pennylane/tape/tapes/tape.py	99.20% <ø> (+0.08%)	⬆️
pennylane/tape/tapes/jacobian_tape.py	98.97% <98.97%> (ø)
pennylane/tape/__init__.py	100.00% <100.00%> (ø)
pennylane/tape/measure.py	96.87% <100.00%> (ø)
pennylane/tape/operation.py	98.33% <100.00%> (ø)
pennylane/tape/qnode.py	98.75% <100.00%> (ø)
... and 4 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 29be042...d0c72a2. Read the comment docs.

[josh146]

💯

To fix the failing Sphinx CI, you'll need to update the

:meth:`~.jacobian`

to be

:meth:`~.JacobianTape.jacobian`

in the QuantumTape.trainable_params docstring

[co9olguy]

Did not review the PR, but if @josh146 approves, go for it!