Social composer: send now and rename messy account names (#6)

## Summary
Closes #5.
- Composer can **Send now** without a schedule time (Schedule and Save draft still available)
- Account names wrap and show the platform id so a bad OAuth label is still readable
- Social accounts page can rename the display name; re-auth keeps a custom name instead of overwriting it with a placeholder
- Worker polls `dispatch_due` every 15s so scheduled campaigns and social posts fire without a manual command; local `docker compose up` now starts the worker

## Test plan
- [ ] Open Social → Compose: confirm **Send now**, **Schedule**, and **Save draft** buttons (no combined Save / publish)
- [ ] Send now with a connected account and caption → post queues immediately, no datetime required
- [ ] Schedule with a datetime → post is scheduled; Schedule with a blank time → validation error
- [ ] Composer account list shows full name plus external id, wrapping if the name is long
- [ ] Social accounts → rename a messy label → new name shows in composer without reconnecting
- [ ] Re-authorize an account that was renamed → custom name is kept
- [ ] Rebuild and start compose (`docker compose up --build`) so the **worker** service is running
- [ ] Schedule a social post ~2 minutes ahead → after due time, worker logs `Enqueued N due item(s).` and the post leaves Scheduled (published or failed)
- [ ] `docker compose logs -f worker` shows `Polling due scheduled work every 15s.`

Reviewed-on: #6

scripts/worker-entrypoint.sh

@@ -47,5 +47,24 @@ uv run python manage.py migrate --noinput
 # worker must too or hashed lookups raise ValueError.
 uv run python manage.py collectstatic --noinput
 
+interval="${DISPATCH_DUE_INTERVAL_SECONDS:-15}"
+echo "Starting dispatch_due poller (every ${interval}s)."
+uv run python manage.py dispatch_due --loop --interval "${interval}" &
+dispatch_pid=$!
+
+cleanup() {
+  echo "Stopping dispatch_due poller..."
+  kill "${dispatch_pid}" 2>/dev/null || true
+  wait "${dispatch_pid}" 2>/dev/null || true
+}
+trap cleanup EXIT INT TERM
+
+# Dev uses ImmediateBackend: enqueue runs inline in dispatch_due. No dj-queue.
+if [[ "${DJANGO_ENV:-}" == "dev" ]]; then
+  echo "Dev worker: dispatch_due loop publishes scheduled work."
+  wait "${dispatch_pid}"
+  exit $?
+fi
+
 # dj-queue supervisor (workers + dispatcher + scheduler). Run on ONE host only.
-exec uv run python manage.py dj_queue
+uv run python manage.py dj_queue