Implement HVM-Style dup

[fogti]

when implementing the superposition stuff can be avoided, imo it should, when it works, we can start to optimize perfomance..., maybe we get fast enough that the not-complete-lazyness of lambda building turns out to be not that much of a problem as expected. Also, I think lambda binders should be implemented via de-Bruijn-indices, which changes the lowering to the language of [1] (because the substitution stuff changes...)
[2] appears to have introduced the superposition stuff, but I'm not entirely sure if it is really necessary.

see also:
[0] https://github.com/Kindelia/HVM/blob/master/HOW.md
[1] https://github.com/zseri/gardswag/tree/main/docs/HVM-without-superposition
[2] https://cs.stackexchange.com/questions/96798/can-a-calculus-have-incremental-copying-and-closed-scopes

[fogti]

Hi @VictorTaelin I'm really sorry to bug you, you're doing awesome stuff, but I don't really understand why the superposition stuff in [0] and [2] is necessary. My proposed alternative would be 1+de-Bruijn, and I don't really get what the obstacle would be which would make implementing it (the alternative) hard, impossible or add additional edge cases or such...