Don't get more than one forkdb read lock in a task

If p2p tries to get a 2nd read lock, write_queue may have asked for a write lock in the meantime, resulting in a deadlock.