Build a cache of files and dirs

Currently, if zip files are included, they get read and stored as a sort of fake file system.

For the non-zip version, it would be useful to do this so that we know what is inside the zipped up files without actually unzipping. This allows for discovery of regional variants. It probably would also speed MathCAT up since it wouldn't need to hit the file system much. I'm not sure about that though because of the cost of reading the zip files. The value of the `CheckRuleFiles` pref needs to be consulted for updating the cache.

Currently the code has a hack/special case for zh/tw. There is an en/gb regional variation, but because "en" gets unzipped for all versions, that case is discovered, but a es/mx variant wouldn't currently be discovered. Also, "zh" is identified as a language directory due to the regional subdir, but since it lacks any files, it shouldn't be. Again, this is only knowable by examining the zip files (currently that bug exists also for the `include-zip` case.

Probably a lot of code used by the `include-zip` case can be reused for building a cache.