WIP: Qiskit 2 support

[MoritzWillmann]

A couple of changes are necessary to support Qiskit 2.0. I started to adapt some things but there are still things missing, also some things wee need to wait on.

• qiskit_algorithms needs to be updated to support 2.0
• BackendProperties doesn't seem to exist anymore and needs to be replaced accordingly in automatic_backend_selection
• Mapomatic needs to support 2.0
• qiskit_machine_learning needs to support 2.0 or remove qiskit_machine_learning dep
• qiskit 2.1.0 drops sympy requirements (Remove sympy as a hard dependency Qiskit/qiskit#14396)

[David-Kreplin]

Maybe a good chance to get rid of the qiskit_algorithms, which is only needed for fidelity quantum circuits.
We have fast implementation for that for PennyLane and Qulacs, but not for Qiskit.

[David-Kreplin]

Mapomatic causes problems with Qiskit 2.0 since it also checks for BackendV1 that is depricated in Qiskit 2.0.
I don't know, if they ever gonna update the Mapomatic Package, maybe we have to create a fix and hope someone will merge it and relase a new Mapomatic version.

[MoritzWillmann]

Mapomatic causes problems with Qiskit 2.0 since it also checks for BackendV1 that is depricated in Qiskit 2.0. I don't know, if they ever gonna update the Mapomatic Package, maybe we have to create a fix and hope someone will merge it and relase a new Mapomatic version.

Should be fixed now (qiskit-community/mapomatic#77)

[github-actions]

This PR is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 1 year.

[MoritzWillmann]

This is mostly done, the only issues persisting are:

• optree derivative using the parameter shift rule shouldn't be defined for non-linear parameters (i.e. if the trainable parameter is within a nonlinear function, e.g. arccos). If optree derivative is called on such a circuit, an error should be raised
• tests for older python versions fail and need to be fixed

@copilot please fix these two things