[TRIAGE]: Investigate whether we need to maintain a "transientModelConfig" for model-config outputs

[mwdchang]

ModelConfiguration operator state is currently defined as

export interface ModelConfigOperationState extends BaseState {
  transientModelConfig: ModelConfiguration;
  notebookHistory: NotebookHistory[];
  hasCodeRun: boolean;
  datasetModelConfigTaskId: string;
  documentModelConfigTaskId: string;
}

Which is then copied whenever we append an operator output. This data is very heavy, especially transientModelConfig which significantly increase the workflow's size and doesn't really like it is being used.

Since output.state is defined as Partial<ModelConfigOperationState> we don't need save every piece of operator.state into output.state.

For example in simulate, we are selective in what output.state gets:

  emit('append-output', {
    type: SimulateCiemssOperation.outputs[0].type,
    label: nodeOutputLabel(props.node, 'Dataset'),
    value: [datasetResult.id],
    state: {
      currentTimespan: state.currentTimespan,
      numSamples: state.numSamples,
      method: state.method,
      summaryId: summaryResponse?.id,
      forecastId: runId
    },
    isSelected: false
  });

See if we can remove transientModelConfig from output.state.

[blanchco]

Since we are now using the side panel to select model configs, we no longer need transientModelConfig to be stored in the outputs array state. We can simply populate knobs.transientModelConfig with either the side panel model configuration or state.transientModelConfig. We can apply the same idea to the intervention policy operator as well since it is using the same pattern