Decouple the MongoDB sinker from the ingester

**Why we need to do that?**

The sink process actually works well, but fails to deliver a controllable performance when the load increases. The goal is to make a controllable streaming where there is no back-pressure. The expected behaviour is to have instance(s) of the sinker that will work in a scalable way (in a way that will be easier in the future to have multiple instances/workers).