Add a regression test for transliterate which happens to match characters

at an EOL/EOF, and therefore should catch the broken behaviour fixed by
Tim J. Robbins in sed(1) recently.

Suggested by:	obrien

tools/regression/usr.bin/sed/regress.sh

@@ -5,5 +5,6 @@ REGRESSION_START($1)
 REGRESSION_TEST(`G', `sed G < regress.in')
 REGRESSION_TEST(`P', `sed P < regress.in')
 REGRESSION_TEST(`psl', `sed \$!g\;P\;D < regress.in')
+REGRESSION_TEST(`y', `sed y/\(\)/{}/ < regress.in')
 
 REGRESSION_END()

tools/regression/usr.bin/sed/regress.y.out

@@ -0,0 +1,4 @@
+input
+data
+for validation
+of sed{1}