* Add two backends: node & web
 * Convert core lib files to use the backends (and not use Buffer)
 * Convert utf16 codec as an example
 * Add testing for both node side and webpack
 * Bump Node.js minimal supported version to 4.5.0 and modernize some
   existing code. This will allow us to get rid of
   safer-buffer, our only dependency.

Three major reasons for reimplementing UTF-16 and not use native codec:
 1. We want to remove StringDecoder & Buffer references due to ashtuchkin#235.
 2. StringDecoder is inconsistent with handling surrogates on Node v6-9
 3. NPM module string_decoder gives strange results when processing chunks -
    it sometimes prepends '\u0000', likely due to a bug.

Performance was and is a major concern here. Decoder shouldn't be affected because it uses
backend methods directly. Encoder is affected due to introducing character-level loop. It's
still very fast (~450Mb/s), so I'm not too worried. If needed, we can make it about 4x faster
in Node.js by introducing a dedicated backend method. Browser speeds will be the same.

…rings.

…uite

To do that I've added a generation step and store the data in test/tables/ folder.