Safely round a floating point number to the nearest integer (Get a stable and portable result when rounding)

TODO:

• Review floating point (sampling and integration periods) to integers(steps and shapes) roundings across simulator components and analyzers.
• Enforce signal periods to be an integral multiple of the integration time step at one place and remove roundings which are only defined in one script (it's offset and error prone)
• Deal with near half integers exceptions, which are the natural points of discontinuity of the nearest integer function.
• Working with powers of 2 seems the natural solution to avoid this floating point issues, and it's currently the way the default simulation parameters are set/ meant to be used.

Read: http://mathworld.wolfram.com/NearestIntegerFunction.html