Previously, we were quoting UNICODE opcode argument with strconv.QuoteToASCII().
However that function, in addition to \u and \U escapes, can produce
e.g. \n, \r, \xAA etc escapes. And all of the latter variants are not
treated as special escapes of a unicode literal by Python, thus leading
to data being wrongly received.

Fix it by doing exactly the same that Python pickle encoder does - the
UNICODE argument comes are "raw-unicode-escape" encoded.

This patch contains only codec tests - not end-to-end pickle tests,
because currently Encoder.encodeUnicode() is called only from under
Encoder.encodeString(), and there only from under

	if e.config.Protocol >= 3

We will indirectly add tests for encodeUnicode @ protocol=0 in the next
patches, while adding support for Python bytes.