Vladimir Ulianitsky

This topic is pretty old, but maybe someone will find this useful: I created [macros](https://github.com/uvlad7/xml2json-rb/blob/0e794a5e5b12632a6a3c69b96147cf5955ff80e0/ext/xml2json/jagnus/src/jni_function.rs) to handle exactly what is described here. It's inspired by [magnus](https://github.com/matsadler/magnus/blob/c2e125e140e1903777ba369f5eabe1b71561d573/src/method.rs) and it's WIP, no...

Кроме асинхронности, было бы неплохо добавить кэширование

@giovanniberti I also added a fix for `Box` conversion here. And I need your advice how to add tests for unchecked conversion too.

I implemented `Object[] Box Box { vec![].into_boxed_slice() } ``` work, because it's not possible to impl foreign trait for slices and it's also not possible to impl `Signature` for `Box`....

@EliseChouleur So, I assume it's better to wait for #59, right?

@EliseChouleur something is really wrong with derived `TryIntoJavaValue` implementation, but I don't see what. UPD: And it crashes in `JavaVM::new(vm_args)`, but only when `User::selfSignatureCheck` call is present. May be [this](https://github.com/jni-rs/jni-sys/issues/26)...

@giovanniberti, so I fixed infinite recursion problem in `[Try]IntoJavaValue`, but only to face another issue with `AutoLocal`s. I see workflows weren't executed, but as for the [22acb1e](https://github.com/giovanniberti/robusta/pull/50/commits/22acb1ebbd2194bdf0dbf5eb3456a6943a8cbeb8), tests fail with...

I assume the problem is in the fact that `user` value is moved and then dropped by try_into - as long as the inner `AutoLocal` - before the actual `env.call_method`...

@giovanniberti , @EliseChouleur , take a look, please. There are still a lot of improvements to be added, but I assume it'd be better to merge this to fix bugs...

And provably it's worth to make `TryIntoJavaValue` [object-safe](https://doc.rust-lang.org/reference/items/traits.html#object-safety), so it'd possible to return dyn values - and arrays of them - to java?