The core of this change is to re-use the escape sequence parsing logic
for parsing both string and character literals.

The actual fix is that UTF-8 encoding was missing for string literals
with \u{...} escape sequences.