Is this the correct place for JavaNano pull requests? Here is some more information:

A while ago, someone from Google (@xfxyjwf?) mentioned that writing to a byte[] instead of ByteBuffer was a conscious design decision for performance reasons. I believe that in an attempt to reduce allocations for utf8 string encoding, byte[] was changed to ByteBuffer in order to interface with code copy-and-pasted from Guava (@charlesmunger). However, if you look into the Guava code, CodedOutputByteBufferNano currently does the following:

1. user provides byte[]
2. byte[] gets wrapped into a ByteBuffer (additional allocation and performance hit)
3. writeStringNoTag(String) calls encode(CharSequence,ByteBuffer) and does some extra work to translate ByteBuffer exceptions
4. encode(CharSequence,ByteBuffer) checks whether it is a direct (0% of the time) or an array backed buffer (100% of the time)
5. encode(CharSequence,ByteBuffer) then extracts the backing byte[]
6. encode(CharSequence,byte[],int,int) writes utf8 into the byte[]

This patch removes the unnecessary steps 2-5 and encodes strings directly into the byte array. The other changes are mostly restored code from before the switch to ByteBuffer.