Test Garph/Gqty as a valid replacement for Pothos/Genql

[ernestoresende]

After recently re-experiencing the GraphQL authoring experience provided by Garph, there is room for some serious reconsideration about having the combination of garph and gqty as a replacement for pothos and genql on scaffolded apps.

The main selling point is the elimination of codegen entirely. Garph ships with built-in type inference based on the declared schema, which can be exported and used in the client on a very similar way compared to what already is done with Genql on the current implementation.

There is an argument to be had in that moving from Pothos means loosing some features like plugins, but to be fair, most of the behavior you would get out of plugins on Pothos is already built-in (subscriptions, pagination, file uploads, and even relay compatible primitives to encourage the usage of the Relay specification).

Past tests with it here shows that switching from Garph to Pothos reduces setup boilerplate (absence of the SchemaBuilder), and simplifies mental model (the API's is just simpler them Pothos's). It also reduces the hindrance of having to run a parallel process for codegen on Turbo.

One thing I'm currently worried about based on reading the examples is how to organize it when things get more complex then just a few queries, altough I know Mitch is already working on that in stepci/garph#78 (I may ask on Discord if he already has a few pointers on recommended setups.