JEP 416 causes StAPI to crash on Java 18 onward

[mineLdiver]

https://openjdk.org/jeps/416

Since all reflection is refactored to use method handles, StAPI's reflection trick to replace final fields no longer works.
This can probably be fixed by using Unsafe instead.

[mineLdiver]

A better approach would probably be analysing the entrypoint classes with SpASM and replacing calls to Null.get() with initialization of the objects required by the class.

However, this won't work with @Entrypoint.Instance static final fields, as the instance is created by Fabric way after the class has initialized. @Entrypoint.Namespace fields won't work either, as we have no way of getting the mod container from a class during its transformation by SpASM.

[Geolykt]

Since all reflection is refactored to use method handles, StAPI's reflection trick to replace final fields no longer works.
This can probably be fixed by using Unsafe instead.

Using Unsafe as a replacement is nonsense too thanks to JEP 8323072.

[mineLdiver]

Well, can still resort to SpASM - by either removing the final modifier from the field's bytecode itself, or generating an initializer.

There's also a cursed approach that involves generating a constructor with SpASM and invoking that through hacky means on an already instantiated object to overwrite values in instance final fields, but that might have already been patched in later Java versions.

[mineLdiver]

At the moment, I've decided to go a safe route:

1. By default, entrypoint utility fields will only be settable if they're public and non-final.
2. Optionally, an entrypoint can expose its private lookup to EntrypointManager in static initializer, allowing EntrypointManager to set private utility fields.
3. Caller-sensitive getters of Namespace and Logger will be introduced as a replacement for final utility fields. Instead of StAPI setting the utility fields on entrypoint discovery, getters will automatically resolve correct values on class initialization. This has proven rather difficult though, so it'll most likely stay marked as "experimental" for some time.

All these changes should make StationAPI fully forwards-compatible with modern Java versions.
There's, of course, EnumFactory, which also has to be dealt with, but I have some ideas for that.