Unicode + add_permastruct breaks rewrite rules

[mikejolley]

This was reported here ​https://github.com/woocommerce/woocommerce/issues/16673

To recreate the issue, create a taxonomy with a cyrillic name such as Сертификат. View the taxonomy archive. You'll see no results; it will go to the homepage rather than an archive.

In WooCommerce you can recreate this by creating an attribute (Product > Attributes) with a cyrillic name, and enabling the 'archive', assigning a term from this taxonomy to a product, and trying to view products by that term.

I managed to trace it back to the add_permastruct. The struct is added with unicode % encoded characters. When the rewrite rules are processed, it thinks these are placeholders so the matches variables do not align. See this screenshot for clarity:

​https://www.dropbox.com/s/5vztnfm6895488a/query%20is%20wrong.png?dl=0

Notice all the $matches? Compare to a working taxonomy:

​https://www.dropbox.com/s/24zyr5v7taw7b60/correct.png?dl=0

This can be fixed by using urldecode when adding the permastruct. I don't know if this has side effects but it worked in testing.

Patch to follow.

[mikejolley]

Fix for 41791

[Soean]

Added a closing parenthes.

[SergeyBiryukov]

Related: #28058

This PR intercepts the permastruct inside WP_Rewrite::add_permastruct() and decodes the URL-encoded slug back into standard UTF-8 text before saving it.
To ensure we don't accidentally corrupt real rewrite tags (where %ca in %category% could be decoded by a naive urldecode), we use a preg_replace_callback that strictly decodes only uppercase hex strings ([0-9A-F]). WordPress's URL encoding always uses uppercase, while built-in rewrite tags use lowercase, making this a safe and robust fix.

Trac ticket: #41791