Provide base inference component

[RobbeSneyders]

By @philippe-ml6:

Implementing GPU components can be a bit tricky since we need to make sure that both preprocessing and inference are batched, would be nice if we can implement a generic inference transform component class that faciltiates this. A bit similar to Beam's run inference API