Fix: Properly rewrite rich text with mixed links

[chosak]

The bulk rich text link rewriter logic introduced in #5875 doesn't work properly if the rich text contains multiple link types where one of those types doesn't need to be rewritten (for example external links like <a href="https://wagtail.org/">).

This commit fixes that bug by handling these cases, and adds additional unit tests to cover the failure cases.

Fixes #11957.

[squash-labs]

Manage this branch in Squash

Test this branch here: https://chosakfixmixed-link-rewriter-q3jbe.squash.io

[gasman]

Thanks @chosak! I see the second issue mentioned on #11957, where a custom LinkHandler is in use, is still present - I think this might be a separate bug, since it involves the replacement tag being written in the wrong place entirely, rather than just URLs being switched around.

[chosak]

Thanks @gasman! I've just pushed a fix and test for that issue as well.

[gasman]

Thanks @chosak! I'm happy with the updated logic, although still concerned that the "zipping tuples of dicts" approach is rather hard to follow, and probably largely responsible for these bugs sneaking in in the first place...

I've now had a go at refactoring it by introducing a TagMatch object to encapsulate all the information for an individual tag - please let me know what you think.

[chosak]

@gasman works for me! This is a nice way to make the logic more straightforward, thank you.

[gasman]

Thanks @chosak! Merged in bf3f87b (main) / 729acab (stable/6.1.x).