[SCHEMA] Reorganize raw dataset rules

[rwblair]

This will be an unpopular reorganizaton. Opening PR to track discussion on other ways we might want to separate out derivatives and raw rules

[tsalo]

Are there any rules that can only be applied in raw datasets, and never in derivative datasets? It feels like any rules labeled "raw" should only apply to raw datasets.

EDIT: If what are currently called "raw" rules can be applied to all datasets, I'd recommend renaming them "general".

EDIT 2: Per Zoom discussion, the application of raw rules to derivative files is complicated. "Derivative" files that are copied from the raw dataset without manipulation follow the same rules as the original files, while derived files use the original file's rules, but shifted down a requirement level (so required metadata fields become recommended, recommended become optional). Implementing all of that is going to be fun!

[rwblair]

Closing in favor of #1288