Improve normalization handling #74
Description
Currently when we save value at given property, flow is as following:
And then at value retrieval again normalization step is considered, (as between store and get the model could have been changed).
Still that appears problematic as we loose precision. e.g. if at some point we want to increase detail of property characteristics. Initially we wanted to save integers, then we decided we need two fraction digits, then we're not able to retrieve that fraction detail from already saved properties as they were already stored as normalized to integers.
Solution would be to introduce only base level normalization at value store (which means we will use normalize from base type, so e.g. will turn strings to numbers for numeric types, but we will not affect number value), and full (corresponding to given property type and definition) normalization at value retrieval.