Unbreak PDF::Content testing

Commit 97d473b hid the conversion of the "\r\n" synthetic
behind the "\r" handling.  Fixed this by moving the check for "\r\n"
forward, so that the speed increase could be maintained.

src/core.c/Str.pm6

@@ -884,11 +884,11 @@ my class Str does Stringy { # declared in BOOTSTRAP
                      ),
                 self!hexify($char),                    # escape since > 0
 #?endif
-                nqp::ifnull(
-                  nqp::atpos($escapes,$ord),
-                  nqp::if(                             # not a known escape(
-                    nqp::iseq_s($char,"\r\n"),
-                    '\r\n',                            # it's the common LF
+                nqp::if(
+                  nqp::iseq_s($char,"\r\n"), # <-- this is a synthetic codepoint
+                  '\r\n',                              # it's the common LF
+                  nqp::ifnull(                         # not a common LF
+                    nqp::atpos($escapes,$ord),
                     nqp::if(
                       nqp::iscclass(nqp::const::CCLASS_PRINTING,$char,0),
                       $char,                           # it's a printable