The i;codepoint collation is the same as the i;octet collation except
that it operates on sequences of Unicode scalar values, not octets.
Note that by definition the set of Unicode scalar values excludes the
surrogate code points and as such they do not occur in valid input.