Fix regex to catch HTML tags

[Wauplin]

This PR fixes the CI for https://github.com/huggingface/course and in particular ./es/chapter5/5.mdx.
Related to this slack thread (internal) and this failing CI.

The failing doc had this form:

10K<n<100K

<Tip>

This is a tip.

</Tip>

in which the _re_lt_html regex detected <n<100K \n<Tip> as a single HTML tag. The _re_lt_html is meant to detect the < characters that are not part of a HTML tag.

I fixed the regex for this use case + added a regression test for it. All previous unit tests are still passing so hopefully it doesn't break something in the wild. I also took the liberty to add verbose mode to explain a bit more what the regex is doing (too me a bit of time to remember 🙄).

cc @mishig25 @MKhalusova @xenova

(also related to #373 and #394 which introduced and modified this regex)

[Wauplin]

CI was failing on an un-related test:

tests/test_autodoc.py:283: AssertionError
=========================== short test summary info ============================
FAILED tests/test_autodoc.py::AutodocTester::test_document_object - AssertionError: '\n<d[221 chars]ters>[{"name": "*args", "val": ""}, {"name": "[462 chars]\n\n' != '\n<d[221 chars]ters>""</parameters></docstring>\n\nBase class[401 chars]\n\n'
Diff is 1304 characters long. Set self.maxDiff to None to see it.
========================= 1 failed, 83 passed in 4.21s =========================

It is due to this commit on transformers that added arguments to the ModelInfo object.

I fixed the expected result in ec14a49.

[MKhalusova]

Thank you for fixing this!

[mishig25]

not to forget this dev change: 8c6163b

[xenova]

Green means go! 🚀🚢 (🟢 for Transformers.js)

• just a reminder to remove that dev change

[mishig25]

lgtm !