Fix bugs that are preventing server from working

[jkomoros]

This broke likely around c6ac49bce0fb7bc16057e668c3357b3ab75052f8

First tracked in #516, some commits have landed against that issue.

The issue was that moves.CurrentPlayer started being executed during manager.SetUp, but it uses games.CurrentPlayerIndex(), which hits CurrentState which hits storage, which doesn't exist. :-/

Tests didn't catch it--it was only when running the server that I noticed an odd error that was preventing it from starting.

0ea579ca98de9c273268f0cc5a5cd7c4d60221c3 is a temporary fix that has mysql storage layer return errors if not corrected, but the problem is more deep.

• [ ] Ensure that go test ./... fails with the issue to ensure future tests will find the issue. In particular, ensure that every example has a test to check that the game can be set up without errors (at the very least)
• [x] Create a StorageWrapper struct that keeps a connected bit, and will auto-fail any call to a getter/setter when connected is not yet called. Take the logic out of mysql, and wrap all storage layers in it so we get the non-connected error for free. IN PROGRESS ON MBP. msyql needs its logic removed, storage.StorageManager needs to grow a CleanUp method (get rid of the test interface?), and iterate until all tests pass)
• [ ] Document which StorageManager's will error if their connect() isn't called eyt
• [x] game.CurrentPlayerIndex() -> state.CurrentPlayerIndex() so there's not an implicit CurrentState fetch. In general, we should make sure all methods explicitly refer to a state so we don't have this confusion in the first place.
• [x] Fix the bug where Blackjack's diagram crashes after create game (which may or may not be related) in EffectiveHand. The problem is that EffectiveHand is getting a sanitized value and not handling it well. It's unclear why it only just now started failing and wasn't broken for a much longer time. I wonder if it's from when I removed the negative-count shadow values a few days ago (024783010d6e2b)

[jkomoros]

Actually thinking about this more, it's weird to have non-trivial logic in a wrapper instead of in the object itself.

Also, Connect() is not in the core boardgame layer. The fact that e.g. memory and bolt don't require a Connect() is a feature, not a bug, because it allows testing in those contexts. We do already exercise the non-trivial behavior (stuff like what caused this bug) in the mysql integration test, which is fine. So this is WAI?

Just fixing the CurrentPlayerINdex() home point is probably sufficient then.

[jkomoros]

I believe these are all now fixed.