properly check for ICU status so known tags get actually recognized

nIcuErr after the second RegexMatcher::find() call was
U_STRING_NOT_TERMINATED_WARNING hence a check for U_ZERO_ERROR failed
and a content string of "foo <emph>bar</emph> baz" became
"foo <emph>bar&lt;/emph&gt; baz" which later provoked a missing ending
tag error.

This should had used U_SUCCESS(nIcuErr) instead and only worked by
chance. Apparently the warning wasn't issued in earlier ICU versions and
only introduced with 55 or 56.

Change-Id: I36482c928544fc6e8c0403997af33084a47d4bea
(cherry picked from commit 84017530)
Reviewed-on: https://gerrit.libreoffice.org/19524Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>

l10ntools/source/xmlparse.cxx

@@ -1144,7 +1144,7 @@ OString XMLUtil::QuotHTML( const OString &rString )
     icu::UnicodeString sReturn;
     int32_t nEndPos = 0;
     int32_t nStartPos = 0;
-    while( aRegexMatcher.find(nStartPos, nIcuErr) && nIcuErr == U_ZERO_ERROR )
+    while( aRegexMatcher.find(nStartPos, nIcuErr) && U_SUCCESS(nIcuErr) )
     {
         nStartPos = aRegexMatcher.start(nIcuErr);
         if ( nEndPos < nStartPos )