Well the Nim compiler includes a virtual machine (compiler/vm.nim) that can interpret Nim code. A "superset" would probably imply to patch Nim's parser/vmgen/vm so that it interprets this "superset", whatever that implies.
But if you think that compilation/interpretation has much to do with how easy it is to add language features, you should re-think your goals as these are mostly orthogonal aspects. At the same time, there are non-obvious interactions between language features. For example, Nim's VM covers a subset of Nim, not a superset.