reduce size of master nodes in the production logging ES cluster

refs: https://gitlab.com/gitlab-com/gl-infra/infrastructure/-/issues/10094

The masters were resized from 4GB to 16GB as part of the issue linked above. Their current CPU utilization is around 15%. The problem we were facing was related to saturation on masters so this should be very carefully examined before any decision is made.

wait for 7.7 upgrade
continue efforts to reduce number of fields and number of shards
collect monitoring data that will further suggest that it's safe to resize the masters