transfer DP methods to the new structure

[[e-mail]]

currently dynamic programming agents are broken. On the long run, this should be fixed by transferring their code to the new structure Will proposed. On the short run, it would be great if we can get the existing code running again.

[Will Dabney]

All the Agents and MDPSolvers should now be working correctly, and have at least one test case in the cases directory. The only exception is KWIK_TBVI, because the comments in the class indicated that it is unfinished. So, KWIK_TBVI has been moved over so it is not broken, and will run without errors, etc, but it is not necessarily 'complete'.

[Will Dabney]

See last comment.