Both /libtracker-sparql/tracker-sparql/tracker_sparql_cursor_next_async
and /libtracker-sparql/tracker-sparql/nb237150 use the main loop.
Let each initialize and run it instead of doing it in main() (yikes!)
or depending on tracker-fts (only true for the former?). Initializing
the main loop on main() was also dubious for the latter, since we fork
into a subprocess.