Implement default values for floats

[OndrejSlamecka]

Before this patch the compiler generates a^b which is not valid Python for floats. A motivation for implementing this (apart from satisfying the capnp specs) are systems for optional values where extremes mark None/Nothing/nullopt/...