Still trying to open buckets when there are pending coordinator packets

[wks]

The assertion error observed in #770 can still be reproduced in rarer cases when GC is triggered very often. One possibility is that GC workers may spuriously wake up when a mutator triggers GC and the coordinator starts executing ScheduleCollection and increments the pending_coordinator_work counter.

Perhaps we should not let GC workers do too much work other than getting executable work packets and executing them.

[wks]

This error can be easily triggered by manually triggering GC (handle_user_collection_request) repeatedly.

The problem goes away if i add a sleep statement (20ms is enough) between consecutive GCs.