Escape supplementary characters for non-UTF charsets #2578

Supplementary characters could be emitted unescaped when serialized with non-UTF, non-ASCII charsets such as ISO-8859-1.

The encoder check used only the low 16 bits, so U+100A3 could be treated as U+00A3 and considered representable, while the original character was then emitted unchanged.

The check now uses the complete code point. This still preserves the existing ASCII and UTF fast paths, with direct and serialization regressions covering the fix.

Fixed by f10c02e