Featured card images have empty alt text

Every featured image renders with an empty alt attribute.

be-best/layouts/_partials/images/featured-image.html:24-25 reads $image.Params.title and .Params.caption, which come from Hugo image metadata. No content defines any: grep '^image:' content returns nothing. The values are therefore always nil, alt="" reaches the markup, and the caption element in image-variants.html:43 never renders. public/post/index.html and public/post/garden/index.html each carry 6 empty alt attributes.

Featured card images have empty alt text

The 44 featured.webp files under content/ each get image metadata: front matter naming the file, with params.title and params.caption. 41 pages gained the block; banana-bread, broccoli-bacon-salad and date-and-walnut-loaf already had one.
When metadata is missing, featured-image.html falls back to the page title, so alt is never empty.
A caption renders under the featured image of a post page, and not under a card.

Acceptance Criteria

grep -c 'alt=""' public/post/index.html and grep -c 'alt=""' public/post/garden/index.html both report 0, against 6 today.
Every <img> written by image-variants.html has an alt equal to the caption in that image’s metadata, and no alt is the empty string on any page in public/.
pnpm run test passes, including html-validate.

Notes

The metadata is Hugo page resource config, the form the three existing recipe pages already used, because the partial reads $image.Params.* off the resource - a top-level image: metadata block would not reach it:
Copy
```
resources:
    - src: 'featured.webp'
      params:
          title: 'Lemon Bars' # caption under the image
          caption: 'A photograph of lemon bars.' # alt text
```
Page layouts (post, recipe, project, profile, about) call the partial with (dict "page" . "caption" true); the four card partials call it with (dict "page" .), so cards get alt text and no caption. Every caller now passes a dict: a page context has no page key to read, and a page errors on the missing field, so a bare . cannot be told from a dict.
Four section images show something other than their section, tracked as p1-is-cuisine-images-show-the-wrong-cuisine, and their alt text says what the file shows: recipe/american and recipe/australian carry the same “Pastries” title card, recipe/mexican carries the “Middle Eastern” one, and recipe/japanese reuses the recipe/chinese beef photo.