Use integer weights

[fyrchik]

float weights are cumbersome. We can use integer ones (or some kind of decimals) as they are easier to work with in multiple languages.

[alexvanin]

I think it is possible to normalize values, considering 1.0 as math.MaxUint64 (32, 16?) and 0.0 as 0.

[fyrchik]

Actually we can use both and corresponding suffix to function names (U64, F64).