Fix parsing larger HTML blocks in MDX files

[mre]

Previously, each HTML line inside a Markdown file would be parsed independently. This worked reasonably well, but in the case where individual lines are not valid HTML, we would skip link checking.

The fix is to accumulate an HTML block until it gets closed and only then parsing it.

Fixes #1923

[mre]

@katrinafyi, @thomas-zahner FYI

[katrinafyi]

Very quick! I'm wondering whether this affects parsing markdown inside HTML, which should work according to the spec? I think this should already be handled by pulldown-cmark, but maybe add a test case with a markdown link inside a html block?

[mre]

Your concern was valid to check, but our implementation handles it correctly. 👍
I've added the test.

Some notes for future reference:

• pulldown_cmark treats block-level HTML tags (<div>, </div>) as separate HTML blocks
• Markdown content between HTML blocks is parsed as normal Markdown paragraphs
• We only accumulate HTML chunks within a single HTML block (between Start(HtmlBlock) and End(HtmlBlock))
• Markdown events like Start(Link) are processed normally and separately

[mre]

Huh, looks like the tests failed, but for an unrelated issue. We should move all tests to "example URLs" at some point and not depend on external servers. 😖

[mre]

Okay, the test fixed itself. My favorite kind of fix. Merging.

[thomas-zahner]

Awesome I see no problems. That was really quick 🚀