gfm parsing oddity with links and raw HTML about commonmark-hs

[jgm]

Minimal repro in try pandoc

[jgm]

As noted in the linked discussion, this only affects parsing with CR+LF line endings.

The issue may be related to #136

[jgm]

Observations:

1. This bug appears with -f gfm but NOT -f commonmark. So it has to do with an extension. Need to isolate which extension with further testing.

2. The bug is in commonmark-hs, not pandoc itself.

   % echo -e "[link](https://baidu.com)aaa<span></span>bbb\n" | commonmark -xgfm
   <p><a href="https://baidu.com">link</a>aaa&lt;span&gt;&lt;/span&gt;bbb</p>
   

3. I can reproduce it even with LF line endings using commonmark-cli, so I'm not sure why things seem different with pandoc.

I will transfer this to commonmark-hs.

[jgm]

Using -xautolinks instead of -xgfm produces the issue. So it can be attributed to the autolinks extension.

[jgm]

The code for the autolinks extension is quite bad and needs work!
There is an extensive set of tests here that we might attend to.
And here is a syntax: https://unifiedjs.com/explore/package/micromark-extension-gfm-autolink-literal/#syntax

Some work in issue147 branch.

[TripleCamera]

Thank you. 🥰