This is a meta issue tracking progress on significantly speeding up loading a schema from disk. There are multiple incremental steps that should get us there.

• Split validation from instantiation. If we have a schema persisted on disk and know that we already validated it before, there's no reason to validate it *again• every time we want to load it (Split schema validation from instantiation #1080).
• Change from classes to simple data objects that are manipulated and queried using functions. This would allow the schema to be persisted and loaded as a plain JSON object.

Note: This is still just a rough outline and there's likely more steps to get there including making sure this is an incremental change and doesn't break existing users unexpectedly.