Optimize Base58 encoding/decoding

[aslesarenko]

Can only be done after #27

Problem

Existing implementation of scorex.util.encode.Base58 is a potential performance bomb, especially encoding, which copy the whole string buffer for each new character (which is added at the beginning of the buffer). It essentially make the algorithm O(n^2) where n is the length of the resulting string.

Solution

• avoid using foldLeft and functional abstractions.
• put characters in ArrayBuilder[Char] and then move them to StringBuilder in reverse order.

[pragmaxim]