feat: implement entry deletion

- database: cascade_cached_bytes_after_subtree_delete — one-pass set-aware
  SQL that excludes the entire subtree simultaneously, avoiding sibling-blob
  cross-counting bug
- database: delete_entry — collects subtree IDs, runs set-aware cascade,
  NULLs archive_run_items.produced_entry_id (FK blocker), deletes children
  then root; ON DELETE CASCADE handles artifacts/tags/collections
- server: DELETE /api/archives/:archive_id/entries/:entry_uid route,
  wrapped in a transaction for atomicity
- frontend: deleteEntry API call, handleEntryDeleted callback in App.jsx
  (optimistic list removal + selection clear), Delete entry button in
  ContextRail with confirm guard
- tests: 4 database tests (unknown uid, subtree removal, run_item nulling,
  cached_bytes recalculation) + 3 route tests (204+gone, 404, 401)

frontend/src/api.js

@@ -58,6 +58,14 @@ export async function removeTag(archiveId, entryUid, tagUid) {
   if (!resp.ok) throw new Error(`Remove failed (${resp.status})`);
 }
 
+export async function deleteEntry(archiveId, entryUid) {
+  const resp = await fetch(
+    `/api/archives/${archiveId}/entries/${entryUid}`,
+    { method: 'DELETE' }
+  );
+  if (!resp.ok) throw new Error(`Delete failed (${resp.status})`);
+}
+
 export async function renameTag(archiveId, tagUid, name) {
   const res = await fetch(
     `/api/archives/${archiveId}/tags/${tagUid}`,