Generalize VibrationalOp.from_polynomial_tensor

[mrossinek]

What should we add?

The VibrationalOp.from_polynomial_tensor implementation currently is a bit more restrictive than it needs to be. This is mostly due to initial design restrictions when the new VibrationalIntegrals class was constructed. Below is an example:

from qiskit_nature.second_q.operators import VibrationalOp, VibrationalIntegrals

d = {
    (0, 1, 2): 2.0,
}

ints = VibrationalIntegrals.from_raw_integrals(d)

op = VibrationalOp.from_polynomial_tensor(ints)

print(op)
# Vibrational Operator
# number modes=1, number modals=[3], number terms=1
#   2.0 * ( +_0_1 -_0_2 )

However, currently there is no way of generating the following VibrationalOp from a PolynomialTensor,
because there is an implicit assumption that the mode index is identical within each pair of creation and annihilation terms:

op = VibrationalOp({"+_0_1 -_1_2": 1.0})

With the introduction of the Tensor class (#1033) and the refactored way of handling the label_template (#1060), it should be possible to also improve the implementation of VibrationalOp.from_polynomial_tensor to be less restrictive.

This would be especially useful for the implementation of #46 and other extensions that might be based on the VibrationalOp.