Add batch_partial implementation

[eddddddy]

Context:
There are scenarios where we want to execute a circuit multiple times, with some parameters identical and other parameters differing between the executions. It makes sense to be able to batch these executions together.

Description of the Change:
Add a new transform qml.batch_partial that behaves similarly to functools.partial, but supports an additional batch dimension in the unevaluated parameters. Now we would be able to achieve the desired behaviour by doing something like:

dev = qml.device("default.qubit", wires=1)

@qml.qnode(dev)
def circuit(x, y):
   qml.RX(x, wires=0)
   qml.RY(y, wires=0)
   return qml.expval(qml.PauliZ(wires=0))

>>> batched_partial_circuit = qml.batch_partial(circuit, x=np.array(np.pi / 2))
>>> y = np.array([0.2, 0.3, 0.4])
>>> batched_partial_circuit(y=y)
tensor([0.69301172, 0.67552491, 0.65128847], requires_grad=True)

More complex scenarios involving callable arguments are also possible. For example,

>>> batched_lambda_circuit = qml.batch_partial(circuit, x=lambda z: z + 0.2, y=lambda z: z ** 2)
>>> z = np.array([0.2, 0.3, 0.4])
>>> batched_lambda_circuit(z)
tensor([0.92032424, 0.87403075, 0.81479384], requires_grad=True)

Benefits:
The user can now add a batch dimension to arguments passed to a partially-evaluated QNode.

Possible Drawbacks:
None

Related GitHub Issues:
Closes #1881

[codecov]

Codecov Report

Merging #2585 (37c2b5b) into master (8265d21) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #2585   +/-   ##
=======================================
  Coverage   99.58%   99.58%           
=======================================
  Files         244      245    +1     
  Lines       19673    19724   +51     
=======================================
+ Hits        19591    19642   +51     
  Misses         82       82           

Impacted Files	Coverage Δ
pennylane/__init__.py	100.00% <ø> (ø)
pennylane/transforms/__init__.py	100.00% <100.00%> (ø)
pennylane/transforms/batch_partial.py	100.00% <100.00%> (ø)

---

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 8265d21...37c2b5b. Read the comment docs.

[antalszava]

Thank you @eddddddy! 🎉 Will be leaving a more detailed review during the first half of tomorrow.

Hi @dwierichs, just a ping to see if the main UI here is looking good. 👍

[dwierichs]

Looking really nice, thanks @eddddddy! Cool addition to support callables as preprocessing :)
Only have some minor confusions/questions regarding the doc string.

[eddddddy]

Hi @dwierichs, thanks for the helpful comments. There are just a couple outstanding that need another look, otherwise I've mostly addressed them.

[antalszava]

Hi @eddddddy, this is looking great overall 🥳 Have a couple of suggestions for refining things, but the overall implementation seems to check out nicely.

[antalszava]

Looks good @eddddddy! 🙂 🎉 A couple of final smaller suggestions and a question.