Actor ended with Actor.fail() should not be restarted

[jancurn]

For context, see the Exit Actor section and this conversation.

Basically, if one uses Actor.fail() to end the Actor run, we should finish the Actor run all the time and not try to restart it. Currently the SDK doesn't support this, and we should make it so - either using an API endpoint to fail the Actor (imo preferred), or some special exit code.

[mtrunkat]

@B4nan , @fnesveda, could you think about this and propose a solution?

I remember that we did not want to use custom exit codes and preferred API endpoints in these cases, as the exit codes could cause problems when using some orchestrators in the future (k8s, etc.).

[jancurn]

+1 to this, API feels cleaner, and also is probably faster than cleanup + process exit