Memory strategies at the whole workflow level

[wlandau]

Related: #19. We should think about when objects should be kept in memory or released. The advantage of keeping them for longer is we do not need to access storage, but that could make memory blow up.

[wlandau]

I guess this could be as simple as the decision to use or not use a centralized cache object. The target's cache object should remain unaffected (we should always clear it after the target builds and stores its value, regardless of the memory strategy). So we won't get to this issue until it's time to implement centralized components. There is nothing we need to implement in the target class.

[wlandau]

New idea: just use the "lookahead" strategy from drake. We can make it efficient by using a second priority queue for downstream targets. For strategies lighter on memory, just use dynamic files.

[wlandau]

Now implemented.