Refactor or better document loaders

[cmungall]

The ways in which ontologies and annotations are loaded is confusing and not well documented.

• everything can be combined into one OWL file, which works for us in monarch but it not intuitive, and requires something like owltools to make the files
• the yaml allows stratification of ontologies and instances TSVs. However, there are 2 kinds of TSVs: associations and labels. It seems only the former is supported
• we need to make this extensible, e.g. to load frequency info for #56
• we have phenopacket loading in a branch by @balhoff but not fully integrated yet?

ultimately we want to be able to fetch latest version from API, to avoid stale files in github

cc @jnguyenx @damiansm @kshefchek