Add public homepage discovery aggregates

Context

The updated signed-out homepage in #87 needs server-owned discovery facts that the current cursor list routes cannot supply honestly.

Needed contract

Add a public, cacheable homepage discovery read that supplies:

• popular bottle links based on a real recent-interest signal;
• the highest-rated bottles for the current year, using the new point-rating aggregate and a server-owned minimum sample;
• the total number of bottles that meet that score sample;
• the number of public critic sources represented in recent reviews;
• the highest-rated qualifying bottle for each most-recorded distillery shown on the homepage.

Each bottle result should use the normal serialized bottle identity and include the point median, point count, and five rating-band counts needed by the existing Score and BandStack components. Each distillery result should keep its normal identity and bottle count, then include the same rating fields for its highest-rated qualifying bottle.

Constraints

• Do not derive popularity or sample-qualified rankings from one cursor page.
• Do not convert critic scores from their native scales.
• Keep this endpoint public and safe to cache.
• Existing recent-review, region, recent-bottle, and global stats endpoints can continue to own the other homepage sections.

Frontend behavior

Until this lands, the redesign will omit the popular and annual-ranking modules and will show the distillery directory without highest-rated bottle details instead of rendering invented values.

Start by tracing the current cursor list routes and the existing recent-review, region, recent-bottle, and global stats endpoints. Define the public cacheable read so it owns the required popularity, annual ranking, critic-source, and distillery aggregates without using one cursor page or converting native scores; done means its bottle and distillery results include the specified identities, rating fields, counts, and sample-qualified totals.