fix: eliminate greenlet crash + memory leak protection

- docker-compose: switch worker to --pool=solo --concurrency=1 --max-tasks-per-child=1
- tasks.py: remove ThreadPoolExecutor wrapper from _run_browser_job (greenlet crash)
- scraper.py: browser fallback runs sequentially instead of ThreadPoolExecutor
- scraper.py: add gc.collect() after scraper close to prevent memory leaks

docker-compose.yml

@@ -144,9 +144,9 @@ services:
     stop_grace_period: 60s
     command: >
       celery -A iaai_scraper.worker.celery_app worker
-      --loglevel=info --concurrency=${CELERY_WORKER_CONCURRENCY:-2} --pool=prefork
+      --loglevel=info --concurrency=1 --pool=solo
       --pidfile=/tmp/celery-worker.pid
-      -Q scraping --max-tasks-per-child=${CELERY_WORKER_MAX_TASKS_PER_CHILD:-3}
+      -Q scraping --max-tasks-per-child=${CELERY_WORKER_MAX_TASKS_PER_CHILD:-1}
     healthcheck:
       test: ["CMD-SHELL", "test -f /tmp/celery-worker.pid && kill -0 $(cat /tmp/celery-worker.pid)"]
       interval: 60s