char8 mode, enabled using the PUGIXML_CHAR8_MODE macro, uses C++20 char8_t
instead of char for the UTF-8 interface. This makes use of pugixml safer
when char is otherwise used for the system codepage.
Stream-based methods received an additional overload, since the char
overload may be used to represent arbitrary bytes, and the char8_t overload
may be used by string streams.
An additional typedef u8char_t, which represents the type pugixml uses for
a UTF-8 code unit, was added for the conversion functions.
Most changes had to be done in the test code. Representing raw bytes as
string literals does not work for UTF-8 literals, since hex escape codes
are interpreted as a Unicode character. Affected places either received a
branch with a u8 literal or use a new RAW() macro which smuggles in UTF-8
code points using chars.
