Deprecate control_wires in ControlledQubitUnitary

[JerryChen97]

Context:
We are in favor of wires, just like what we did with MultiControlledX.

!Note: this involves a big user-interface change. With PL version <=0.40, for ControlledQubitUnitary, we have arg wires to be the target_wires, and the arg control_wires is a positional arg. To be consistent with all the other interfaces, we would like to alter this into wires to be positional and to represent control_wires+target_wires

Before:

mat = np.eye(2)
control_wires = [0]
wires = [1]
ControlledQubitUnitary(mat, control_wires, wires=wires, control_values=...)

After:

mat = np.eye(2)
wires = [0, 1]
ControlledQubitUnitary(mat, wires, control_values=...)

where the new wires should be the concatenate of old control_wires and wires

Description of the Change:
Besides the modified test files across the whole codebase involving ControlledQubitUnitary, we made the following necessary adjustments beyond the __init__ method:

• classmethod _unflatten, _primitive_bind_call and method _controlled were adjusted accordingly following the new interface
• move the TypeError regarding empty arg wires to the beginning.
• add a register for ControlledQubitUnitary in bind_new_parameters.py following the pattern of CRX
• modified the direct calls of ControlledQubitUnitary inside src files metric_tensor.py, controlled.py and matrix_ops.py.

Eco-System

• Catalyst: frontend/test/pytest/device/test_decomposition.py Adjust the usage of ControlledQubitUnitary catalyst#1483

• Lightning: PR to fix it update usages of ControlledQubitUnitary in tests pennylane-lightning#1047

• • test_gates.py
• • test_gates_and_expval.py
• • test_measurements_class.py

• QML Demos: tutorial_classically_boosted_vqe.py; PR to fix it Fix controlled qubit unitary usage qml#1310

Plugins

• Pennylane-AQT: No instances of deprecated code found.
• Pennylane-Qiskit: No instances of deprecated code found.
• Pennylane-IonQ: No instances of deprecated code found.
• Pennylane-Qrack: No instances of deprecated code found.
• Pennylane-Cirq: No instances of deprecated code found.
• Pennylane-Qulacs: No instances of deprecated code found.

Benefits:
Unified interfaces with other controlled ops.

Possible Drawbacks:
This is a hard change to deprecate this arg. Would have to introduce errors for testing matrices.

Related GitHub Issues:
[sc-80842]

[github-actions]

Hello. You may have forgotten to update the changelog!
Please edit doc/releases/changelog-dev.md with:

• A one-to-two sentence description of the change. You may include a small working example for new features.
• A link back to this PR.
• Your name (or GitHub username) in the contributors section.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.49%. Comparing base (666941f) to head (980135f).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6839      +/-   ##
==========================================
- Coverage   99.54%   99.49%   -0.06%     
==========================================
  Files         477      477              
  Lines       45246    45331      +85     
==========================================
+ Hits        45041    45102      +61     
- Misses        205      229      +24     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[astralcai]

I feel that specifying wires with base being an operator is going to be ambiguous regardless of what we do. I believe we're deprecating that soon as well?

[JerryChen97]

Indeed. Since this PR is altering way more than the other I did this first. Anything comes here is temprory and handwaving just to pass existing usage.
My plan for deprecating Operator input is to take the matrix and wires out from the op along with raising warning, and then fall back to normal process.