LayerdEndodingCircuit in QSVC with PQK not working in cross_val_score

[JanSchnabel]

Describe the bug
I am initializing an empty LayerdEncodingCircuit object and pass it to a PQK, which I pass to a QSVC. Then I want to set the encoding circuit using set_params(encoding_circuit_str=fm_str) and put the resulting QML model into scikit-learns cross_val_score. This gives me the error

TypeError: LayeredEncodingCircuit.init() missing 1 required positional argument: 'num_qubits'

To Reproduce
I am using the develop branch of sQUlearn. This is a minimal example to reproduce the error

from sklearn.model_selection import train_test_split
from sklearn.datasets import make_classification
from sklearn.model_selection import cross_val_score

from squlearn import Executor
from squlearn.encoding_circuit.layered_encoding_circuit import LayeredEncodingCircuit
from squlearn.kernel import ProjectedQuantumKernel, QSVC


x,y = make_classification(n_samples=100, n_features=8)
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=42)

num_qubits = 8

lfm = LayeredEncodingCircuit(num_qubits=num_qubits)
quantum_kernel = ProjectedQuantumKernel(
    encoding_circuit=lfm,
    executor=Executor(),
    parameter_seed=0,
    gamma=0.5
)

svc_kwargs = {"C": 100}
qsvc_model = QSVC(
    quantum_kernel=quantum_kernel,
    **svc_kwargs
)

fm_str = 'Rx(p;=0*p+np.pi,{p})-Ry(x;=np.arctan(x),{x})-'
qsvc_model.set_params(encoding_circuit_str=fm_str)

cv_score = cross_val_score(qsvc_model, x_train, y_train, cv=5, scoring="roc_auc")

Expected behavior
Before merging the "removing num_features" branch into develop it worked.

[MoritzWillmann]

On it