Send errors upstream (to client) and terminate

[tskluzac]

We need a means of terminating (and notifying users) for the following reasons in the crawler_worker:

• bad Globus Auth credentials
• misspelled or incorrect file paths
• any GlobusError that has been retried n (=10?) times without success.
• SQS errors that have been retried n (=10?) times without success.

I think this update should happen via the crawl-worker updating the postgres entry for the crawl by having it update status to 'error' and ALSO add a descriptive free text 'message' column (that if things are running fine could just be something like 'ok') that can describe reason for error.