admission: integrate token-based AC into overload dashboard

[tbg]

Related to #154449. We should touch up the dbconsole metrics to include the new metrics to enable troubleshooting of this feature. In particular, we should think about how to surface which tenants are noisy, which goes beyond what metrics can natively express (at least using internal timeseries). We could expose a "most consumed tokens per node" gauge whose value is a tenantID - it's a crutch but might work. SREs might have opinions on what is really needed here.

Epic: CRDB-54929

Jira issue: CRDB-55040