Add Decimal64.valueOf() methods which specify expected precision

BGPCEP is a rather heavy user of 'type decimal64', which is now being mapped to yang.common.Decimal64.

It is instantiating Decimal64 values from either a float or a long. In both cases the model defines that fraction-digits are equal to two and there is some weirdness around it as well.

Introduce a few methods to predictably create these values.