fix updates of floating point data with interval null values

[Dimitar Misev]

There are two problems when updating floating point data with interval null values:

1. The floating point null value is wrongly cast to int64
2. If the source value is null then the target value is not updated, but it should be if the target value is also a null value