Default "excerpts" don't take HTML entities into account

[janboddez]

See https://indieweb.social/@janboddez/111075331718609311

[janboddez]

What happens is we first parse the template tags in a status, and then decode entities. But we run mb_substr() on the encoded version.

We gotta still decode at the end, though, because we fallback to the title if the parsed template is an empty string. Or we gotta decode that title before falling back to it.

[janboddez]

OK, that seems to work.