Add call in RunInference API

Right now beam.ml.inference.api.RunInference does not call anything and is an empty API.

It should call base.RunInference.

We should also investigate changing the api to work more like a transform.

Right now it might look like this:

api.RunInference(pcollection, modelLoader)

but we might want it to follow other transforms and have it run like this:

pcollection | api.RunInference(modelLoader)