[Idea] Better memory handleling in encoding process!

[OmidHekayati]

Due to the nature of JSON, it is easy to predict(estimate) needed memory in the marshaling process and it is a much more easy process in the code generation process. But your implemented idea about multi buffer and combine them finally is not efficient enough now to win the battle with the standard language package! We work on the JSON package with this idea and the idea works very well and just needs to improve to handle more features in the code generation part.

Read more about idea easily by read our benchmark test here: https://github.com/SabzCity/libgo/blob/master/json/benchmark1_test.go#L285

See very basic benchmark here: https://github.com/SabzCity/libgo/blob/master/json/benchmark1_test.go#L20