Draft: QNN refactoring

[David-Kreplin]

Branch for refactoring the low-level QNN implementation to make it more modular:

• [x] Create QNN Base Class derived from the current low-level QNN implementation (@David-Kreplin)
• [x] Adjust existing low-level QNN program to the Base Class (@David-Kreplin )
• [ ] Modifie Loss functions to return Loss and Gradient computed from the QNN(@MoritzWillmann)
• [ ] Adapt training.py to the changes in the Loss function (@MoritzWillmann)
• [ ] Add x=QuantumCircuit support to the existing low-level QNN implementation (@MHagelueken and @David-Kreplin)