* Use more reasonable buckets for `process_msg` histogram * Another adjustments of the buckets * Move historgram timer to a more relevant place * Add a dedicated collation distribution time metric * Cargo fmt