Fast bootstrapping of patches

[nmlgc]

This would only download the most necessary data of each patch (game version data in particular) on initial patch selection, with the rest being only downloaded when needed (e.g. once a particular game has been located on a user's system).

In particular, this means:

1. Only pull versions.js for every patch after stack configuration
2. Use that to start searching for games
3. Download any global patch data during the search process
4. Create shortcuts
5. Download/update the rest of each game's data individually after the user started a game through a shortcut, using a blocking downloader.

[nmlgc]

So I've been looking at the implementation of this, and it revealed a slight flaw in the current patch file structure: <game>.js and <game>.<build>.js are stored in the root directory of a patch. The patches themselves don't include a simple list of the games they support (and really, why should they), so determining the global and game-specific parts of a patch becomes slightly less trivial, as there is no way around building such a list internally by parsing through files.js.

If we are about to break backwards compatibility anyway, we should move these files into the subdirectories of the respective games (so that e.g. th14.v1.00b.js becomes th14/th14.v1.00b.js) for even faster bootstrapping. Then, global files are easily identified by the lack of a slash in their filename.

For now, I'll be implementing this feature as if that were the case. In a typical setup where only a single game is to be patched, this would still download 40-ish JSON files too many.