Separation of errors

In fact, there are two types of errors occurring in Mistral:
* "Expected" errors related to particular workflows. An example of that is an error returned by std.http action (timeout error or error status like 500).
* Errors caused by malfunctions of Mistral components. For example, if Mistral DB goes down it will be this kind of error. It could be conventionally called an infrastructural error.

This separation is very important in the context of analysis a particular workflow execution. These two types of errors should be exposed differently from API perspective and in logs. When analysing logs we should clearly see whether an error came from workflow or from Mistral component failure.

REQUIRES SPEC