Serialization performance

[eric-millin]

JsonSerializationWriter looks like it could be a good deal more efficient. With a small change I improved performance by about 50% and the tests pass. Those numbers come from benchmarks, so they only mean so much. I can PR the change at some point if you're interested, if not, no worries.

eric-millin@1366790

Here are the differences in the benchmarks

Current   	  111204	     10472 ns/op	   20806 B/op	     211 allocs/op
AfterChange 	  170420	      6276 ns/op	    2095 B/op	     100 allocs/op

[baywet]

Thanks for the suggestion. We'd be very happy to see that performance improvement in the code-base assuming it doesn't trade-off functionality.

[eric-millin]

We're not really impacted by this issue. Perhaps I'll do it on a Friday afternoon.