Optimize images hosted on mindtouch files even if they return a "bad" application/octet-stream mime type

[benoit74]

Sample:

[mindtouch2zim::Thread-5 (worker)::2024-11-26 18:31:57,143] DEBUG:Not optimizing, unsupported mime type: application/octet-stream for asset from https://geo.libretexts.org/@api/deki/files/5324/xenolith-of-diorite.jpg?revision=1 used by page ID 7820 (https://geo.libretexts.org/Bookshelves/Geology/Physical_Geology_(Earle)/08%3A_Measuring_Geological_Time/8.02%3A_Relative_Dating_Methods), page ID 30611 (https://geo.libretexts.org/Courses/Coalinga_College/Introduction_to_Earth_Science_(C-ID%3A_GEOL_121)/02%3A_The_Geosphere/2.27%3A_Measuring_Geological_Time-_Relative_Dating_Methods)

This represent a very high number of assets (e.g. 4135 out of 23357 for geo.libretexts.org), so we need to fix this.

[benoit74]

We should probably try to guess mimetype based on URL when mimetype is application/octet-stream:

>>> mimetypes.guess_type(urlsplit("https://geo.libretexts.org/@api/deki/files/5324/xenolith-of-diorite.jpg?revision=1").path)
('image/jpeg', None)

Note that we need the urlsplit to retrieve the path, guess_type does not work well with URL including a query parameter