Incorrect regexp character class handling with \uNNNN and \xNN code points

[akosthekiss]

Test code:

function re_test (pattern, string)
{
  var r = new RegExp (pattern);
  print ("'" + pattern + "' ~ '" + string + "' => '" + (new RegExp (pattern)).exec(string) + "'");
}

re_test ("[\\u0020]", "u");
re_test ("[\\u0020]", " ");
re_test ("[\\u0020]", "x");

re_test ("[\\x20]", "u");
re_test ("[\\x20]", " ");
re_test ("[\\x20]", "x");

Output:

'[\u0020]' ~ 'u' => 'u'
'[\u0020]' ~ ' ' => ' '
'[\u0020]' ~ 'x' => 'null'
'[\x20]' ~ 'u' => 'null'
'[\x20]' ~ ' ' => ' '
'[\x20]' ~ 'x' => 'x'

The first and last results are incorrect.

This could either be fixed in #961 or handled separately (in which case a regtest could be added to the test suite as well, based on the above).

[akosthekiss]

It seems that a ch = RE_CHAR_UNDEF; is missing from
https://github.com/Samsung/jerryscript/blob/master/jerry-core/parser/regexp/re-parser.c#L424
and
https://github.com/Samsung/jerryscript/blob/master/jerry-core/parser/regexp/re-parser.c#L436