AggregationPipeline.reinterpret doesn't actually work

When using .reinterpret (or any function that calls it internally, such as .group), the Java driver still tries to deserialize the initial document.

Can be fixed by specifying the expected output document in JvmMongoCollection.kt:322. We should keep track of the output KClass<*> and KType during the creation of the pipeline (in PipelineChainLink?), so it is available there.