Improve performance of WAD search

[Phytolizer]

Basic idea:

• On first run, build a KV store from "base" names to actual file paths. e.g. "doom2.wad -> iwad/doom2.wad"
• On subsequent runs, check the database first before crawling the potentially large directory structure.
• If the name is not found in the database, crawl.
• If the name is found in the crawl, rebuild the database.
• If it is not found, give up.

With this basic flow, commonly used WADs will take next to no time to load. When adding WADs, it will only be slow the first time.
The worst case scenario is when a WAD is new, either we will crawl twice or simply save the results of crawling as it goes, then check for presence and update the DB later.