Better shutdown behavior on MIGRATION event about crawlee

[jancurn]

Also, AutoscaledPool should probably stop launching new tasks. And when all tasks are finished, it could just exit the process with an error exit code, so that the task is migrated to new server.

[jancurn]

BTW if actor doesn't have "Restart on error" flag set, then we shouldn't send any migration event, since the actor won't be able to handle it anyway

[mtrunkat]

I think that those are two different cases. "Restart on error" means that error in actor is something that may occasionally happen as, for example, memory exceeded when doing big long crawl.

Some of the actors for example - actor that aggregates data from a dataset and sends email is something I want to restart on migration but not on error - because error should not happen there.

[jancurn]

You're right, this should probably be other setting - e.g. "Fail on migration"

[mnmkng]

Do this once #195 is done.

[mnmkng]

Implemented in #268