Add an unpatcher to revert the monkeypatch, mid-run?

[cchance27]

For people using the nodes for part of the workflow would be good if you added an UnApply Pulid Flux ...

Basically just revert the application, drop the pulid_ca, etc, and set forward back to forward_orig's target.

if not hasattr(flux_model, "pulid_ca"):
            # Add perceiver attention, variables and current node data (weight, embedding, sigma_start, sigma_end)
            # The pulid_data is stored in Dict by unique node index,
            # so we can chain multiple ApplyPulidFlux nodes!
            flux_model.pulid_ca = pulid_flux.pulid_ca
            flux_model.pulid_double_interval = pulid_flux.double_interval
            flux_model.pulid_single_interval = pulid_flux.single_interval
            flux_model.pulid_data = {}
            # Replace model forward_orig with our own
            new_method = forward_orig.__get__(flux_model, flux_model.__class__)
            setattr(flux_model, 'forward_orig', new_method)