Minimal Ractor support

[rmosolgo]

I've been following along with @byroot's recent Ractor-related blog posts, and it makes me want to make GraphQL-Ruby play nice with non-main Ractors. I think it will be a long road ... but it starts with a single step! I'm opening this PR to start a conversation (if anyone is interested) about how/what to accomplish.

Goals for non-main Ractors:

• Execute a query and return a result Hash to the main Ractor
• ~~Parse schemas from definition~~ Not currently possible because Schemas are classes and non-main Ractors can't assign instance variables of classes.
• Somehow provide an escape hatch for fetching data. For example, could the dataloader interface work for sending IDs up to the main ractor, then receiving data in the non-main Ractor? (Ractor-friendliness in GraphQL-Ruby is meaningless if there's now way to make an ActiveRecord query during execution...)
• Stretch goal, maybe someday: build a GraphQL::Query in the main Ractor, then run a dependent Query::Partial in a non-main Ractor. This is very hard because GraphQL::Query has a bunch of dynamic caches.

Non-goals:

• Modifying schema definition in a non-main Ractor is impossible

Currently, GraphQL-Ruby supports incremental, dynamic schema definition: types, fields, etc, can all be added on-the-fly, modifying schema definitions. But I think Ractor-friendliness will require some API for freezing a schema definition. RactorShareable does that here.

---

TODO:

• [ ] Support looking up types and fields in non-main Ractor
• [ ] Try a Partial in a non-main Ractor
• Another PR: implement pass-through dataloader