Support arxiv html papers

[dai-shuo]

Arxiv provides static html version of most papers using LateXML. The html contents are well structured by rich ltx_xxxx CSS classnames. It should be lightning fast parsing those paper htmls and get very precise info.
It would be cool to support arxiv html parsing, as a much faster branch or a strong hint for the pipeline.

[PeterStaar-IBM]

@dai-shuo Excellent point! In my latest PR (#240), this is indeed possible,

For example, if you run this,

poetry run docling --from html --to md "https://arxiv.org/html/2408.09869v3" --output ./scratch/

you will get,