Make database format more language-agnostic

[loafofpiecrust]

There are a few areas of the back-end that we assume only Cherokee data will be represented.
We can start re-engineering some of that code to allow for extension into other languages.

• Layers of linguistic annotation should be an extensible enumeration (enum AnnotationLayer) rather than simply optional fields (phonemic: Option<String>). This encourages a cleaner architecture where we add new variants to AnnotationLayer instead of more fields to AnnotatedWord. It also allows the data to define a specific order of layers, though that may not be desirable anyhow because generally layers of linguistic analysis are presented in a specific static order.
• Phonetic and morpheme layers may be returned in several orthographies, which are Cherokee specific.