New entrypoint which runs available detectors based on configuration

[wandgitlabbot]

In GitLab, by Daniel Oosterwijk on 2019-11-29

• It should be possible to dynamically produce a dataflow where several detectors are running in parallel.
• Ideally, it would allow us to have a single source and sink of each type that needs to be created, and feed their input/output flows to the detectors that need it.
• We should allow for external configuration of this flow, essentially boiling down the complex Flink graph to a simpler configuration file that just sets inputs and outputs for each detector.
• Chaining custom ProcessFunctions (including detectors) would be useful, as well as including arbitrary generic functions (map, filter, etc) at any point in the pipeline.
• Detector configuration should still work as it does currently, where a simple external configuration file overrides the built-in defaults.
• The format of the custom flow configuration file should be investigated, since if we want it to be the same file as the detector configuration, it will need to be in the same format. We are open to changing the format of the detector configuration file.

[wandgitlabbot]

In GitLab, by Daniel Oosterwijk on 2019-12-03

eaacca54 added the UnifiedRunner, which achieves most goals of this issue. Detectors can be toggled on or off, but their dataflow is defined in code and is not configurable via external file.

Closing this issue until the need for more complex, arbitrary-at-runtime dataflow paths arises.