reduce number of threaded workers

context: the igwn-alert-overseer uses more memory than the old lvalert-overseer. Under heavy load where many alerts were being sent out, python was unable to allocate more memory for multiprocessing threads to send out alerts. This reduces the number of threaded workers to keep under the current memory limit.

The long-term solution is to move away from multiprocessing for sending alerts, which is blocking gunicorn's async worker type. the igwn-alert overseer should be modified to just take in a list of topic names and then send them out on their own.

there's also a cert/account update.