Add finished_at timestamp to datasets

[dale-wahl]

super simple read last line from log. only using it in the dataset results details at the moment.

using it for the sort feature is not trivial unfortunately since we are reading from the log instead of a database column.

possibly it would be better to use something like this in a migrate script to update the database instead? actually, i immediately like that idea more. tell me why not.

[dale-wahl]

Switched to added finished_at timestamp that is created when DataSet.finish() is called. Created migrate script to fill in finished_at for existing datasets using the last log entry. In testing on my existing 4CAT instances, I did find examples of datasets without logs as well as some logs where the final line does not contain a timestamp.

The non-existant log files were from either a failed 4CAT import (the 4CAT import changes the dataset key and, somehow, lost the log file in the failure) and a deprecated analysis that left a hanging dataset (weird dev case I think). The unable to parse were all related to failed datasets that were marked as failures but with long errors as the final log entry.

[stijn-uva]

This is kind of nitpicky, but could we change the column name to timestamp_finished? I think we use timestamp_something for all columns containing timestamps (in any case I've tried to do that), and consistency is nice.

Looks good to me otherwise.

[dale-wahl]

Changed to timestamp_finished.

This PR has a migrate script and thus needs a VERSION change. Do we auto run migrate on restarts now? If not we'll want to release on merge.