Event logging in case of error

[jankatins]

Two problems in this PR:

• No slack message when the mara DB is unreachable ("Notify when the runlogger.handle_event() throws an exception" // "Make event handling robust against exceptions during handling" and "Make the commandline only succeed if a RunFinished event arrived" make it more robust in general)
• Old error messages in slack when used in the browser and using multiple runs with errors without restarting flask ("Clean the message cache when a pipeline run starts and finishes")

The first showed up when the mara DB was shut down unexpectedly and we didn't get a slack notification (until our alerting based on last finish times kicked in). the second when you do SQL only fixes which need no restart...

[martin-loetzsch]

It looks good, but why do we need a new Event type? Couldn't we just send Output events? It looks to that the event is only used for printing / displaying stuff.

[jankatins]

Will remove.

When I wrote this I wasn't sure if the code path has a valid pipeline. Seems it does and this exception will probably also not be reused outside of pipelines runs...