Simplify anchorize()

xml, sourcepos

EOF here means we bail out, not continue

Also go a little further on XML.

In the cmark-gfm AST, "tasklist"s actually replace list items, rather
than being nested within t...

Implement Default for enums without using #[default] attribute

Add quadratic fuzz target

The fuzzer can be run with `cargo fuzz run quadratic`.

optimise plugins

This can really add up! We also weren't lock()ing stdout, which is a
big miss.

Fixes https://github.com/kivikakk/comrak/security/dependabot/8.

fuzzing, and fixes from it

Upstream relies on there being a trailing NUL; we do not have the
luxury.

cleanup: tasklist, front matter

* We only want to try parsing front matter once, not on every feed.
* Replace regex with somethi...

store/expose String in AST, not Vec<u8>

I've added some unsafe from_utf8s where it took no effort at all to
recognise that it would alwa...

You can't truncate the String after like you can a Vec<u8>; the char
boundary check will complai...

refmap: port github/cmark-gfm@b4f01062

Address quadratic complexity issues from upstream, pest

Maintain a linebuffer between all calls, not just the last one and
finalise.

It's obviously unnecessary; just superstitiously wondered why it was
still there in upstream, 8 ...

NULs are not allowed in the input stream (which re2c handily enforces
for us), so we can return ...

A few specs to fix up now.

The design of Comrak's autolink module only vaguely matches that of
upstream's, but we can still...

This fixes the regression back up.

Fixes the quadratic behaviour, but one regression test fails. More
ports to come.

Removes quadratic behaviour in number of successful parses in one line.

Adapted from core::slice::ascii::trim_ascii_start/end. I thought
perhaps the quadratic behaviou...

We were still doing the manual_scan_link_url even when the spacechars
scan had failed.