Incomplete parsing on bad attribute data

[Frenzie]

To reproduce:

Take a user guide that has a stray quotation mark:

<img loading="lazy" src="https://hosted.weblate.org/widget/koreader/user-guide/multi-auto.svg" alt="Translation status"" style="width:60%;">

Run something like:

pandoc user_guide/en.html --resource-path=.:user_guide:user_guide/pictures --metadata title="KOReader User Guide" --toc --toc-depth=3 -o koreader_user_guide.epub

And voila, this nonsense:

<img src="../media/file1.svgz" style="width:60%;" loading="lazy" data-"="" alt="Translation status" />

Parsing stops after that.

Now this is probably more of a Pandoc bug and obviously an issue with the original HTML as well, but the reason I'm writing it down is that it's merrily ignored by at least some other apps.

koreader_user_guide.zip

I don't think it's a bug, but occasionally you still want to act on these kinds of things. :-)