enh: support multiple types of prediction

[isabelizimm]

resolves #2

allows users to pass in "predict", "predict_proba", or "predict_log_proba" to a vetiver_post() call in order to create another endpoint that uses that type of prediction for scikit-learn models.

# example
api = VetiverAPI()
api.vetiver_post(endpoint_fx="predict_proba")