Improve performance of byte handling

[danielcompton]

There may be more efficient ways to handle the bytes coming and going from RethinkDB than creating and parsing as strings. A few possibilities

• Look at gloss :char-sequence on the codecs, and if it actually improves performance. We need to handle two types if we go this route, a java.nio.CharBuffer, and a :gloss.data.string.core/char-sequence (both are CharSequences, but the gloss one doesn't give access to the underlying array).
• Look at using Netty JSON decoding instead
• Look at using io.netty.handler.codec.protobuf
• Reusing any resources
• Parsing the custom RethinkDB types in the JSON parsing

[jwr]

To encourage this: I've been looking at performance problems with clj-rethinkdb recently and from my profiling it appears that gloss.io/decode_stream is a major performance problem.

Additionally, gloss seems to have been abandoned.