design(search): render results as a dense ruled list, not floating cards

[johnpc]

What

Search results rendered each hit as a full-width floating board card (border + drop-shadow + radius, title stacked above its meta line, padding: 14px, gap: 8px). In a single-column result list that meant tall rows with content clustered on the left and a large empty right region on desktop, and airy, low-density rows on mobile — each result eating ~180px for a title + one chip.

Search was the last core task-list surface still using floating cards after My Tasks (#220), the Calendar agenda (#221), and the List view (#109) were made dense. This brings it in line, scoped to .search__hit:

• Strip the card's border / box-shadow / border-radius / surface fill; give each row a single border-bottom hairline (--tf-border) + a --tf-surface-2 hover highlight, and zero the .search__results gap so the dividers butt into one continuous ruled list.
• Lay the row out horizontally — title on the left (ellipsized), meta cluster (project chip / due label / assignee avatar) aligned to the trailing edge — matching the My Tasks / Calendar / Search-consistent row rhythm. The title truncates so a long name never pushes the meta off-row.

Token/CSS-only; no markup, logic, or schema change (the DOM is unchanged, so the existing SearchHit tests are unaffected).

Before / After

Desktop — results read as a scannable list; no empty right region.

Before	After

Mobile

Before	After

This completes the dense-ruled-row treatment across every core task-list surface (List, My Tasks, Calendar agenda, Search).