Input: read and validate full UTF-8, both Basic Multilingual Plane
and extended characters.
Collate surrogate pairs as specified in RFC467.
This ensures that UTF-8 strings that reach the application are
always valid.
Output: Assume UTF-8 strings provided for output are valid.
The escaping was broken, fix this by not encoding UTF-8 characters with \u.
Simply write them to the output stream as-is.
This is the right thing to do. See https://www.ietf.org/rfc/rfc4627.txt:
"JSON text SHALL be encoded in Unicode. The default encoding is UTF-8."
Fixes#16.
Also add tests for the new functionality.