Convenience function to add charge perturbations with Nedeljkovic Soref Mashanovich model

[marc-flex]

This PR adds a convenience class that allows to easily specify medium perturbation using the model in here. It addresses issue https://github.com/flexcompute/tidy3d-core/issues/846

For instance, in the Charge notebook one can now define the perturbation medium like this

perturbation_model = td.NedeljkovicSorefMashanovich(ref_freq=freq0)

si_perturb = td.PerturbationMedium.from_unperturbed(
    medium=si_non_perturb,
    perturbation_spec=td.IndexPerturbation(
        delta_n=perturbation_model.delta_n(),
        delta_k=perturbation_model.delta_k(),
        freq=freq0,
    )
)

The class NedeljkovicSorefMashanovich derives from a new class AbstractDeltaModel so potentially more perturbation models could be implemented following this approach (or used as template).

[dbochkov-flexcompute]

not sure what would be the best organization of this kind of model, given it so specific to a single material. I guess, one alternative I see is to simply have something like IndexPerturbation.from_nsm_si_model(f: float) class method instead of a new class. One concern, neither that nor the current implementation easily allow to include heat perturbations too

[marc-flex]

I agree this model is very specific. That's why I made it derive from a more generic base class. This base class, in principle would allow to implement other models for other use cases. It would also allow users to create their own and use them with

def from_perturbation_delta_model(cls, deltas_model: AbstractDeltaModel) -> IndexPerturbation:
    """Create an IndexPerturbation from a DeltaPerturbationModel."""
   return IndexPerturbation(delta_n=deltas_model.delta_n, delta_k=deltas_model.delta_k)

Precisely because of it being so specific I'm not super keen on having something like IndexPerturbation.from_nsm_si_model(f: float). I'd prefer to have the specific functions somewhere else.

[momchil-flex]

I think I agree with @marc-flex here, @dbochkov-flexcompute if you find a quick moment to give final thoughts if we can merge this as is?

[dbochkov-flexcompute]

yeah, sounds good!