Fix surrogates parsing on regex

[jedel1043]

Just a simple fix to ensure unpaired surrogates are correctly handled instead of ignored. Also modified the range check for surrogates, since we just want to check if the first surrogate is high; if it is low, it is already unpaired, so we can directly return.

[jedel1043]

can you please share an example regexp which did the wrong thing before?

Both new tests are examples of regular expressions that failed before the change, but some simple examples would be any succession of two unpaired surrogates:

Regex::new(r"(?:[\uD800\uDBFF])").unwrap_err();
Regex::new(r"(?:[\uDC00\uDFFF])").unwrap_err();

[ridiculousfish]

Makes sense, thank you! Merged.