add a reload/update icon in Graph settings (Left Node Actions) to test the async route of the backend

Overall what happened is: ngrams table was changed (new table version). On Graph GET, versions were compared and an expensive Louvain clustering recomputation was triggered (if versions differed). This caused API slowdown. Idea is: remove that version comparison on GET (i.e. serve whatever the latest computed graph version is). On graph change (via "remove candidate/remove stop"), do it into the /api/v1.0/graph/{graph_id}/async endpoint, wait for job id to finish, then rerender graph. The job will recompute graph in the background.

Batch ngrams table change: https://drive.google.com/file/d/1O0MTkULK4U1VRGzSddyKUL0Gxk8ItJ2s/view