Hardening: sanitize agent input

- if agent sent logs -> is the step currently in created/pending/running -> allow it else deny
- if workflow report as finished mark all still running steps as not running (statevis for debate)
- make sure workflow who is finished can not be "reopend"
- ...