Flush logs before status is complete about girder_worker

[zachmullen]

This line in __exit__ takes care of ensuring logs are flushed once the task is complete.

[manthey]

Right, but the status is changed BEFORE the logs are flushed. Specifically, if you are listening for girder events where the log is updated, you can get a job whose status is completed but logs aren't flushed. This was surprising, as in a test I expected an error message to be in the logs when the error status was set (but it wasn't there yet).

[manthey]

I would recommend adding self._flush() in updateStatus before the status request is made.

[zachmullen]

I'm fine with that change.

[manthey]

Added in PR #65.