Issue/1828 aggregate ref with convertable

[schauder]

This resolves the issue at hand, but in an ad hoc way, which feels very wrong. I still feel like I/we are doing something wrong with these conversions.

The problem here is:

Given an AggregateReference<E, C> with E an entity and C a type for which a write converter is regsitered, we currently determine C as the target type of the conversion.
With this information we call the AggregateReference converter, which extracts the id value and determines it is done, because the result is of the desired target type.

This is demonstrated by the failing new unit test.

The integration test is fixed by just slapping a getPotentiallyConvertedSimpleWrite on the already converted type, which looks exactly for the customconversions, that we currently missing.

But this means we do 2 custom conversions. Why 2? Why not arbitrary many?

[mp911de]

But this means we do 2 custom conversions. Why 2? Why not arbitrary many?

It should be exactly one to apply ConversionService. Converting AggregateReference via ConversionService from AggregateReference<S, …> to AggregateReference<T, …> isn't what we're looking for. Instead, we want to store the reference and using its identifier. Implicitly applying conversion, also without further context of identifiers creates a big ball of mud that obfuscates the actual processing.

Instead, we should be explicit that we work with a reference and apply domain-specific rules of the underlying entity.

For the time being, we can adopt the fix to align with the remainder of conversion. However, we should refactor AggregateReference into a clean approach.

[schauder]

I rebased the PR with the intend to clean it up. But it turns out the solution collides with the resolution of #1842