utf8 encoder speedup for large amounts of data processing

[antelle]

Thanks for work!
I have tried to use your package on nodejs to encode/decode 50MB data files but it has consumed too much memory and failed with allocation error. In this pull request, I have replaced arrays with stream processing without double memory consumption and made strings buildup with chunks which increases speed and doesn't bloat V8 GC.
Now it's converting pretty much faster. Also I've added performance test for large amount of data.

[mathiasbynens]

Thanks for this patch!

Could you please:

• respect the existing coding style, i.e. don’t change tabs to spaces
• don’t bump version numbers
• make sure the code coverage % remains stable?

Thanks.

[antelle]

Okay, I'll fix it and resubmit a new pull request tonight.