Restore type_for(type) function

[antoine-malliarakis]

Motivation

Following #1622 the function Tapioca::Dsl::Helpers::GraphqlTypeHelper#type_for required an GraphQL::Schema::Argument (instead of a GraphQL::Schema::Wrapper) to be passed.

Though this did make perfectly sense, we actually used that method to explicitly compute types in our DSL process without this being necessarily "for a field". For example it could be for union types.

Implementation

What I did was:

1. Rework the type_for method to be backward compatible with the pre Update GraphQL mutation compiler to consider arguments with replace_null_with_default #1622 code (e.g. keep accepting a sole GraphQL::Schema::Wrapper argument as a parameter)
2. Introduce an explicit type_for_argument which would be redirecting appropriately to keep Update GraphQL mutation compiler to consider arguments with replace_null_with_default #1622 feature

Tests

I added tests for some introduced adherence.

[antoine-malliarakis]

I could have put

• a simple T.class_of(GraphQL::Schema::Member)
• a T.untyped

But, the first option is referring to a "private API" which I'm not really fond of, and the second option seems to be missing the point...

[KaanOzkan]

Implementation looks good and thank you for the tests.

I'm not 100% sure if we should support this but I'm leaning towards yes. I'll wait for another reviewer to confirm it's okay before merging.

[antoine-malliarakis]

Flagging as draft as there were other refactorings which made this PR actually outdated. Will rework.

[antoine-malliarakis]

@KaanOzkan rebased at head to take into account latest modifications. Not too impacting it seems.

@vinistock do you think you could have a look at this PR eventually ? (Or is there someone else I would need to contact to that extent ?)

[vinistock]

Changes look good to me. There are some CI failures though

[antoine-malliarakis]

Changes look good to me. There are some CI failures though

I believe I fixed the linting (or more specifically, the type check error). I'm not too sure about how to reproduce the errors reported for the other ruby versions though. Still checking.

[KaanOzkan]

Failures were unrelated to this PR