Custom SQL for most heavy queries

[kdembler]

While the squid by itself brings a huge performance difference compared to the QN (10-15 minutes -> 30 sec) when fetching all the objects that a given distributor should serve, 30 sec at 100% CPU usage is still not a great result considering we fetch "only" 1M objects. My guess is that the auto-generated SQL query is quite unoptimized. Since getting all objects for a given worker will be done quite often, there's a lot of space for optimization. I think we could write a custom query with hand-written SQL that ensures good performance for those big queries.