[RFC] Respect nullability type hint in response handling

[kuhnroyal]

Currently we execute requests with dio.get<T>() and the response data is always of type T?.

In the future we could respect the type hint and return response data of type T.

1. dio.get<T>() => T
2. dio.get<T?>() => T?

For case 1 we would have to add error handling for the special case of the response unexpectedly being empty.

This would be a breaking change and thus would require a major version bump.

This was first raised in #1292

[kuhnroyal]

@wendux @jgoyvaerts What do you guys think?

[AlexV525]

Even we strict the generic type to T, the T itself extends Object? by default. For somehow it's not helping the nullability, but can be a improvement of static analysis.

[kuhnroyal]

Yea that's true.

[kuhnroyal]

Actually that is not really true. We can provide the correct response type. This is a breaking change but we could target this for a 6.0 release: #1755