* Fix 64-bit modbus sensor register reads
When reading four 16-bit modbus registers as a sensor value,
slave output is stored first as 64-bit integer, but before returning
that value is converted to double precision floating point. This
causes rounding errors for integer values bigger than 2^53.
After this change floating point conversion is done only if user
has configured scaling or offset using floating points.
* Formatting
* Review fixes
Bas Nijholt
Tomi Lehto