Use UTF-8 as the base of conversions

[viktoriafordos]

As the built-in support of Erlang has shifted to UTF-8 from latin1, it would be a good idea to follow this tendency by altering the code base of ejson, too.

[jonasrichard]

Atoms and string are handled as unicode (utf-8 encoded data), binaries are as-is, numeric and boolean things should be not considered here, so I think it is completed.