Deterministic IntegerMath

[MaxGraey]

@vgrichina Currently there are pretty basic concept of IntegerMath without overflow checking yet. All others functions will be stubs with "not supported" exceptions. WDYT?

Relate to #465

DONE

• tests

сс @Jannis

[vgrichina]

looks good overall

[dcodeIO]

LGTM, but wondering whether IntegerMath does quite cut it compared to let's name it SafeMath that also works with floats but does NaN canonicalization (or whatever is necessary to make it deterministic). IntegerMath feels a bit like it's most useful for games, which is great, but then it might or might not cross the line for what should be included with the standard library and what should become a userland module? Other than that I'd absolutely use it if I had a use case, OK to merge for me, but mentioning. Any thoughts?

[MaxGraey]

@dcodeIO NaN canonicalization should be done on virtual machine side because this would be faster, safer and don't increase binary size.

[vgrichina]

@dcodeIO NaN canonicalization should be done on virtual machine side because this would be faster, safer and don't increase binary size.

@MaxGraey this is also better for us cause e.g. in distributed consensus context we cannot trust that contract is compiled to canonicalize NaNs, unless it happens either on VM or post-processing stage (vs on compile stage).

[MaxGraey]

I will implement this in different way. But current road block is #1137.

Closing for now due to general cleanup described here #1277.