[Feature] Create a server-level consumer rate limit

follow up on #6291. we now have a way to throttle on a per-table basis. but we don't have a way to limit the global rate for the server.

I don't have a good idea on how to limit on a server level or how to distribute the rate limit fairly to each realtime table. but this would be a good feature to have to safeguard, otherwise ops will have to come up with clever configuration balance for each table.

Start by reviewing issue #6291 and the existing per-table throttling implementation referenced in this issue. Define how a server-wide consumer rate should be enforced and fairly distributed across realtime tables, then document measurable behavior for the global limit and its interaction with table-level limits.

Written by the indexing model from the issue text.