Allow ability to run processors in parallel if defined as such

[MKLeb]

Something like this:

my_pipeline:
  - collect: my_collector
  - process:
    - proc_1
    - parallel:
      - proc_2
      - proc_3
    - proc_4
  - forward: my_forwarder

This way, we could get the full benefit of a DAG, Apache Airflow style (https://en.wikipedia.org/wiki/Directed_acyclic_graph). These parallel steps would theoretically interact with orthogonal subsets of the data, so there would be no concerns of race conditions. We would need to figure out a way to gather the event after the parallel steps in such a way where we only have one event with the combined, processed data. Will most definitely require changes in the pipeline manager.