Serialization version numbers heavily dependent on CH version

Serializer version number right now is tied to the CH version number. This is wrong as in most cases the serialization data model won't change as often as CH versions. Need to introduce simpler, counted version numbers for serialization data.