Raise max_worker_processes on start-up if it is too low

We sometimes see systems with an insufficient number of background worker slots for Citus operations (incl. starting the maintenance daemon, master_update_node). This most likely happens because parallel workers use up all available slots.

The default configuration of PostgreSQL (max_worker_processes = 8, max_parallel_workers = 8) does not leave enough room for extensions to create background workers on a system that runs parallel queries.

If possible, we should raise max_worker_processes on start-up in the same way we raise max_prepared_transactions.