Improve the query schema builder

[garrensmith]

The current query schema and the builder are a direct port from the original scala code. Given the teams improved knowledge of rust and how we are using the schema, we can improve on the original design and make it a little more rust-like.

At this point I don't have a very clear path on how we would do, but some of the ideas I have are:

• Move the schema into its own crate.
• Implement something very similar to the walker pattern for the data model work where each type has its own walker for example the EnumWalker.
• Remove the arc's

An approach to implementing this could be something like:

1. Chat to the query engine team for general ideas and thoughts on a new design
2. Implement a very rough POC that tests out some of the ideas we would want to implement
3. Depending on the POC either rewrite the schema builder in one go or start rewriting smaller parts of it in the new design

[tomhoule]

Not in my domain but I just wanted to chime in on the walker pattern. We've been using it more and more in migrations team parts of the codebase, and we've refined it in the process. The best example (and the most recent) IMO is the walkers from parser-database. They're used heavily in the validations in datamodel core and the general opinion is that it's much nicer to work with than previous solutions.

[tomhoule]

Public API for the parser database walkers: https://prisma.github.io/prisma-engines/doc/parser_database/walkers/index.html