Distributed Labeling Function

- Do we want to parallel compute the prediction engineering process?
- Does the API in dask allow for a data slice generator?

### Process

Initial thoughts on what the process might look like:
- A dask data slice would be input to the labeling function.
- The graph of the dask data slice would be extended (not computed) by the labeling function.
- The label times would be computed by iteration based on a search criteria or persisted on the cluster.