Having logs stored in redis then in database took too much memory in redis and
storage in database.
Using files is far simplier, but it require to have a shared file system (nfs)
when the worker and the cubiweb instance (the reader) are not in the same
server.

Use the new cw_celerytask_helpers filelogger instead of redislogger.
Logs are stored in celerytask-log-dir directory in gzip with a predictible
filename based on task_id (which is unique).

Drop task_logs attribute from CeleryTask and update tests accordingly.
celery-monitor don't copy anymore from redis to database when the task is
ended.