AST access by resolvers

[alloy]

Some field resolvers need the AST in order to execute. Always inlining all the AST for each field will presumably get expensive in terms of bundle size n(needs verification). Possible solutions:

• Normalize ASTs for all combined queries so that sub-trees can be shared.
• Only include AST for fields that were defined with a directive that instructs the compiler what AST will be needed. (E.g. also some of its sub-fields)
• Statically inferring what fields might be needed is not trivial, but might be feasible with e.g. strict TS code.

[freiksenet]

I see two possible approaches to this. One is handling AST as an async chunk that is either loaded manually by the user or automatically when info is used. Another is to limit what you can have in info object - possibly we can say that you can't directly access AST (or schema) when using this framework.