Fix category chips: dead links, and the wrong category on the card

[ssavutu]

Two defects behind Janine's screenshot of a men's lacrosse article showing under Columns. The second is what she reported; the first I found alongside it and is live right now.

1. Chips link to pages that don't exist

The chip is a link, and its target was derived from the category name alone via CanonicalizeSlug, independent of site_taxonomy:

/mens-basketball    200   ← the real subsection page
/men-s-basketball   404   ← what the chip linked to

Both basketballs and both soccers — the four apostrophe subsections — have had dead chips this whole time. Same apostrophe possessiveVariant already absorbs on the matching side; the link never got the same treatment. Nobody reported it because a chip is a small target and the article underneath works.

Now resolved through site_taxonomy, falling back to the derived slug so the value is never empty. A category that only reaches a section by alias — men's lacrosse after #199 — links to that section instead of nowhere.

Two things to review:

• RefreshCategoryAliases loads every section and subsection now, not just rows with aliases, because a row with no alias still has to resolve. That's most of them.
• A row's own title outranks any alias. Otherwise Roll sports with no subsection row back into Sports #199's "Men's Basketball" alias on sports would have sent the basketball chip to /sports instead of its own page. A chip should reach the most specific page listing the article.

Tags are deliberately untouched: no taxonomy rows, so derived slugs are all they have.

2. The card shows the wrong category

The card takes categories[0]. That article is ["Men's Lacrosse", "From the Playbook", "Sports"] — it's in Columns because of From the Playbook, but it announced itself as a sport.

Ordered in the handler rather than taught to the card: the handler is the only place that knows which section was asked for, so doing it there fixes every surface at once — homepage blocks, section pages, subsection pages, filtered listings — with no component changes.

The two fixes are coupled, which is why they're one PR: matching a category against the section's slugs only works once the category carries the slug the section knows it by. Fix 1 is what makes fix 2 possible.

Rules:

• A subsection stands on its own — on a subsection page the subsection is the reason the article is there, not the parent it also carries.
• No section context, no reordering — author pages and unfiltered feeds keep the article's own order, since no category is more relevant there. Single articles untouched for the same reason.
• Stable partition — two categories both in the section keep the article's order between them, so the kicker doesn't move for unrelated reasons.

Testing

go build, go vet, full suite green. The database integration tests were run against a real MariaDB 11.7 rather than left to skip on a missing CMS_TEST_DSN, including the new alias-vs-own-page precedence case.

One thing the unit tests can't cover: a possessive category only resolves to its slug through the taxonomy cache, which lives in the database package and can't be loaded from a handlers test. The ordering tests use Field Hockey instead, and the comment says why.

🤖 Generated with Claude Code