Low level emulation of floating point values and arithmetic

[Washi1337]

Summary:
Concrete floating point values are currently not masked with a known bits vector. The concept of (partially) unknown floating point values are therefore not yet known in the emulation framework. The proposal is to adjust the implementation of Float32Value and Float64Value to a more low level implementation, and to support low level arithmetic operations on them as well that takes the unknown bits into account.