pgmq.metrics() can slow down when queue size is large

[ChuckHend]

metrics function calcs the min and max of each message here. There is no index on enqueued_at, and we dont need an index there for normal read() and send(), since we sort by msg_id. However, getting the max and min of enqueued_at can slow down when there are millions of messages in the queue. A temporary fix could be to create an index on enqueued_at there will regularly be large number of messages in queue.

[haydenflinner]

Perfect use-case for a BRIN Index, would have as close to trivial overhead as I think is easy to do in pg.