DB-Level Changefeed: Add per-table scope to lag metrics behind a cluster setting

DB level changefeeds should have the ability to scope metrics to each table that they watch, but that feature should be off by default to prevent cardinality explosion. Put this feature behind a cluster setting.

Epic: [CRDB-1421](https://cockroachlabs.atlassian.net/browse/CRDB-1421)

Jira issue: CRDB-51247

Start by locating the DB-level changefeed lag metrics and the cluster-setting definitions; no files or tests are named in the issue. Confirm how per-table scope should be enabled and what default-off behavior and metric output should look like, then add coverage for the setting and scoped metrics.

Written by the indexing model from the issue text.