Be able to load/save named snapshots

Currently the snapshot feature only stores the one most recent snapshot per task. There are many uses cases where you would want to keep several versions around and be able to reference which snapshot should be used.

One use case is that a task contains a UDF and uses the snapshot to store parameters about its learned model. Maybe you want to try out a new model but do not want to loose the existing working model. Having a mechanism to save the existing model and refer back to it later is crucial.

I am not sure how this is going to work yet, but I am thinking we need to add some kind of large file store to Kapacitor and it will probably be an immutable store that creates versions instead of overwriting existing data. The limiting the number of versions kept plus marking a version as permanent.

Start by locating Kapacitor's current snapshot save and load implementation and any related tests; the issue names no files or entry points. Clarify how named versions, retention limits, permanent versions, and large-file storage should work before defining acceptance criteria for saving and selecting multiple snapshots.

Written by the indexing model from the issue text.