TODO

For 1.0:

- When running the execute loop, the results of store lookup should be cached. If 100 tasks are all dependent on
    this one unfinished initial task, then we shouldn't query the store 100 times. [Dec 3 2009]

- Think about how to use weak references to keep the results luke-warm [Dec 3 2009]

- Time tasks and display them with the status. This is actually very easy to implement with redis (with assynchronous
  increment operations), but not so much with a file based store, so it might be 'optional' depending on the backend.
  This could be used to give an estimate of how much time is necessary to finish [Dec 3 2009]

- For every jugfile, save a dump of the tasks and their hashes [Jan 6 2010]

Sometime:

- Write an SQL backend [Dec 13 2009]

- Unload leaf-task results but not others [Dec 28 2009]